diff --git a/README.md b/README.md
index 24825a9..3dd0e55 100644
--- a/README.md
+++ b/README.md
@@ -244,7 +244,7 @@ What each retired piece used to provide, and what covers it now:
 | Retired piece | Covered by |
 |---------------|------------|
 | `androidtv` playback state | the ADB media-session state ladder (`playing` / `paused` / `buffering` / `idle` / `off`, plus `on` only when no source reported a power or playback fact) |
-| `androidtv` app name / `app_id` lookups | the manager's store + the app-name map + the manifest label of the app that is currently on screen (fetched on demand; a failed fetch is retried on the next poll instead of being cached as a permanent miss) |
+| `androidtv` app name / `app_id` lookups | the manager's store + the one app-name chain: your apps map, then the device's own label for the app on screen (fetched on demand; a failed fetch is retried on the next poll instead of being cached as a permanent miss), then the built-in app names, then the package id |
 | `androidtv_remote` power, volume and key commands | the protocol transport: the `remote` entity, `media_player` volume, `KEYCODE_*` commands |
 | `androidtv_remote` availability | the availability contract (protocol `is_on`, ADB awake, power switch) |
 | `universal` media player templates | the media player entity's own attributes plus the `media_dag_*` attributes |
@@ -391,11 +391,16 @@ The **Apps** row opens a small editor rather than one long form. Home Assistant
 *Edit* and *Remove* are offered only while there is something to act on: a table with
 no remembered rows and no discovery yet shows just the text field and *Add an app*. The
 editor's description lists every app it knows — the rows you saved, and the ones
-discovery found, marked `[detected]`. **Edit** acts on both (that is the point: the
-Shield exposes no app labels of its own, so a detected app reads as `name (package id)`
-or as `package id (no name yet)`, and picking it is how you give it a name — saving
-stores it as one of your own rows). **Remove** acts only on rows you saved, because a
-detected app has no line to delete (it would simply come back with the next sweep).
+discovery found, marked `[detected]`. **Edit** acts on both (that is the point: an app
+the device does not name reads as `Your Name (package id)` when the map or the built-in
+app names know it, and as `package id (no name yet)` when nothing does — picking it is how
+you give it a name, and saving stores it as one of your own rows). **Remove** acts only on
+rows you saved, because a detected app has no line to delete (it would simply come back
+with the next sweep).
+
+A discovered app that nothing can name is **never pre-filled into the text field** as
+`package=package`: a row like that would store the package id as if it were a name. It is
+offered in the row editor instead, where "no name yet" says what it is.
 Removing a row never uninstalls anything from the Shield — the app simply stops being
 offered. Rows you do not touch are left exactly as they were.
 
@@ -508,10 +513,10 @@ outcome that leaves you guessing.
 ### App discovery (what is actually installed)
 
 The Configure page can ask the Shield which apps it has, so the **Apps** table does not
-start from a four-entry guess. Pick **Discover installed apps** on the Configure menu and
-the table is rebuilt from two ADB queries — and the row then **says what happened** (see
-*What the Discover row tells you* below), because an action that appears to do nothing is
-indistinguishable from a broken integration:
+start from the built-in app names alone. Pick **Discover installed apps** on the Configure
+menu and the table is rebuilt from two ADB queries — and the row then **says what happened**
+(see *What the Discover row tells you* below), because an action that appears to do nothing
+is indistinguishable from a broken integration:
 
 | Command | What it returns | How it is treated |
 | --- | --- | --- |
@@ -529,15 +534,44 @@ capture: no real Shield was available when this was written, so
 fixtures — the same "inferred, not observed" mark the state=6 buffering value
 carries in the source ladder above.)*
 
-Names come from the same cached path the Current App sensor uses: a `pm dump` per
-package, parsed for `application-label`, cached for 24 h. A full sweep runs the dumps
-in a small burst (4 at a time, 50 ms apart), which finishes a typical Shield in
-seconds instead of the minutes a strictly serialised sweep would take; the commands
-share the watchdog's own ADB connection, and `adb-shell` serialises the I/O on it, so
-the burst queues politely instead of interleaving on the socket. A label fetch
-that **fails is not cached**, so that app shows its package id in this table and the
-next sweep tries again — this is the same rule as the entity path, and it is why a
-transient ADB hiccup no longer sticks.
+Names come from the same cached path the Current App sensor uses, and there is ONE
+chain behind every surface — this table, `media_player.app_name`/`source`, the Current
+App sensor and the Apps editor. In order, the first answer wins:
+
+1. **your Apps map** — your own edit always wins;
+2. **the device's own label** for that app, read from `pm dump <package>`: the probe
+   greps **on the device** for the label shapes Android prints — `application-label:`,
+   `nonLocalizedLabel=` and `labelRes=` — so only those lines cross ADB (a real dump on
+   one reported Shield is ~2241 lines of dumpsys, and the old `head -300` cut it before
+   the package section even began, which is why it looked like "no labels");
+   a quoted value counts, `null` and a resource id like `0x7f0f0d31` do not;
+3. **the built-in app names** (see *Sources* below) — a curated, community-maintained
+   `package → name` table;
+4. **the package id**, shown **once** (never as `pkg=pkg` in a row).
+
+A label probe is cached for 24 h, and its outcome is remembered either way: a probe that
+**ran** and found nothing is a definitive negative (zero further round trips for that
+app), while a probe that **could not run** is not cached at all and is retried on the next
+poll — this is the same rule as the entity path, and it is why a transient ADB hiccup no
+longer sticks.
+
+A full sweep runs a handful of probes at a time (4 at a time, 50 ms apart), which
+finishes a typical Shield in seconds instead of the minutes a strictly serialised sweep
+would take; the commands share the watchdog's own ADB connection, and `adb-shell`
+serialises the I/O on it, so the burst queues politely instead of interleaving on the
+socket.
+
+**Some devices expose no labels at all.** The reported Shield is one: nothing in that
+2241-line dump names any app. So after 5 probes in a row come back "ran, no label" the
+integration stops spending a shell per package for the rest of the 24 h TTL: the sweep
+answers the remaining packages from the map instead of probing them, names come from
+tiers 1, 3 and 4, and a later device change or update is re-checked when the TTL
+expires. The Configure menu says so in its verdict when it happens, and the table's
+`name` is then your map's name, the built-in name, or **empty** — an app nothing can name
+keeps the id in `package` alone rather than inventing a name out of it.
+
+An app that nothing names is still offered by the **Apps** row editor (it reads
+"`com.example.app` (no name yet)"), which is where you give it one.
 
 Icons are **best effort and explicitly heuristic**: `pm dump` does not expose an app's
 icon resource (an extracted APK's icon is a bitmap only its own code can resolve), so
@@ -563,6 +597,7 @@ one-line verdict, which replaces the usual "not discovered yet" note:
 | --- | --- |
 | `N apps found on the Shield — the table is updated.` | the sweep ran, and the editor's text field now carries the discovered rows |
 | `the Shield reported no installed apps — the table keeps your saved rows.` | the sweep ran and found nothing to add; your rows are untouched |
+| `N apps found — this Shield exposes no app labels over ADB, so names come from the built-in app names and your own rows.` | the sweep ran, and the device itself named nothing: the per-package probes have stopped (see *App discovery* above) |
 | `the discovery sweep could not reach the Shield — check the ADB connection; your saved rows are unchanged.` | the sweep failed; nothing was cached as a result and nothing you wrote was lost |
 | `ADB is not set up for this entry — nothing could be discovered.` | this entry has no ADB manager (a remote-protocol-only setup), so there is nothing to sweep |
 
@@ -588,11 +623,20 @@ sweep could not run), while the Configure row forces one.
 ### Sources (the app list in a dropdown)
 
 The media player's source list is your **Apps** map — the same map app-name resolution
-uses. With no map configured it falls back to a small built-in list
-(TiviMate, Wholphin, Netflix, YouTube), so a universal-media-player card always has
-something to show. Discovery fills the **Configure** table, not this fallback: the
-dropdown stays exactly the map you saved, so what a card offers is always what you
-chose. An app that discovery found and you want in the dropdown is one row to add.
+uses. With no map configured it falls back to the built-in app names, which on a device
+that exposes no labels *is* the naming: the table covers the apps the reported Shield has
+(Projectivy Launcher, Samba (SMB), X-plore, Plex Media Server, TiviMate, Hyperion Grabber,
+Prime Video, MediaSession2MQTT, Speech Services by Google, Button Mapper, Wholphin,
+Findroid, Pluto TV, YouTube, Downloader, TV Quick Actions, Argosy, Netflix, RetroArch and
+the Google TV launcher, plus Google's speech services), so a card shows
+"Netflix" instead of `com.netflix.ninja` with nothing configured at all.
+
+That table is a **community-maintained convenience, not device truth**: package ids are
+stable, but display names change with vendors and app updates, and an app it has never
+heard of falls through to its package id. Your own map overrides any row in it — one line
+in **Apps** — and discovery fills the **Configure** table, not this fallback, so the
+dropdown stays exactly what you saved. An app that discovery found and you want in the
+dropdown is one row to add.
 
 **Apps are sources, not browser media.** This integration no longer publishes the app
 directory through HA's media **Browser**: the Browse dialog shows an **empty tree** for
@@ -610,9 +654,11 @@ entity: media_player.living_room_shield
 ```
 
 - **`source_list`** — the app names. They come from the map (in the order it lists
-  them), so editing the map renames the dropdown entries.
-- **`source`** — the app currently on screen, resolved exactly like the Current App
-  sensor: your map first, then the cached device label, then the raw package.
+  them), so editing the map renames the dropdown entries; with no map, the built-in
+  app names above.
+- **`source`** — the app currently on screen, resolved by the same chain the Current App
+  sensor uses: your map first, then the cached device label, then the built-in app name,
+  then the raw package id.
 - **`media_player.select_source`** — picks one **by name** and launches it over the
   protocol, the same path `play_media` uses for apps. A raw package id works too (that
   holds the package id), names match case-insensitively, and a name that
@@ -640,7 +686,8 @@ The map is your own override, so two shapes are accepted rather than papered ove
 - A **padded value** typed by hand (`com.x=  Netflix  `) is shown cleaned in the
   dropdown, but `app_name`, `source` and the Current App sensor
   display the map value **verbatim** — so `source` can be a string that is not in
-  `source_list`. A blank value falls through to the raw package instead.
+  `source_list`. A blank value falls through the rest of the chain instead: the device
+  label, then the built-in app name, then the raw package id.
 
 ### Remote command script
 
diff --git a/custom_components/nvidia_shield/config_flow.py b/custom_components/nvidia_shield/config_flow.py
index d6d5771..e3d879f 100644
--- a/custom_components/nvidia_shield/config_flow.py
+++ b/custom_components/nvidia_shield/config_flow.py
@@ -154,11 +154,15 @@ def _apps_table_text(
     """The Editable apps table: stored rows first, then discovered apps (t57).
 
     Every app the Shield reports gets a row, so the page opens on the REAL app
-    list instead of the four curated entries ("it doesn't find all the apps").
+    list instead of only the built-in names ("it doesn't find all the apps").
     The stored rows keep their position, name and icon — the user's edit is
     never overwritten by a discovery — and a discovered app the map does not
-    mention is appended with the device label (or its package id when the label
-    fetch did not answer) and its heuristic icon.
+    mention is appended with the name the SAME chain resolves (user map > device
+    label > curated map) and its heuristic icon. An app whose name is unknown
+    anywhere is NOT written as ``pkg=pkg``: a row that saves the package id as if
+    it were a name is the defect the user reported from this very field, so such
+    an app is offered in the row editor (which says "no name yet" out loud)
+    instead of being pre-filled here.
 
     The icon column is emitted only for a stored icon, because the discovered
     icon is an MDI NAME (``mdi:netflix``) and the icon column's documented
@@ -178,7 +182,11 @@ def _apps_table_text(
         if not package or package in seen:
             continue
         seen.add(package)
-        name = app.get("name") or package
+        name = str(app.get("name") or "").strip()
+        if not name or name == package:
+            # nothing names this app: the id would be doubled into a "name" that
+            # is not one (t66). The row editor still offers it for naming.
+            continue
         lines.append(f"{package}={name}")
     return "\n".join(lines)
 
@@ -1287,6 +1295,13 @@ class NvidiaShieldOptionsFlow(config_entries.OptionsFlowWithReload):
                 "the Shield reported no installed apps — the table keeps your "
                 "saved rows."
             )
+        if getattr(manager, "app_labels_absent", False):
+            # Measured, not guessed: the sweep probed and this firmware named
+            # nothing, so say which names the user is looking at (t66).
+            return (
+                f"{len(apps)} apps found — this Shield exposes no app labels over "
+                "ADB, so names come from the built-in app names and your own rows."
+            )
         return f"{len(apps)} apps found on the Shield — the table is updated."
 
     async def async_step_menu_again(self, user_input=None):
@@ -1530,19 +1545,24 @@ class NvidiaShieldOptionsFlow(config_entries.OptionsFlowWithReload):
         """The apps the device reported, ``{package: {name, icon}}`` (t64).
 
         Straight from the t57 cache — no sweep, no shell. The name may be empty:
-        the user's Shield exposes no device label at all, and a package whose
-        ``pm dump`` failed falls back to its id, which is NOT a name (see
-        ``_editable_apps``).
+        the user's Shield exposes no device label at all (t66: the discovery table
+        no longer repeats the package id as a name), and an app nothing names is
+        left empty so ``_editable_apps`` can mark it as needing a name.
         """
         detected: dict = {}
         for app in self._discovered_apps():
             package = str(app.get("package") or "").strip()
             if not package:
                 continue
+            name = str(app.get("name") or "").strip()
+            if name == package:
+                # a name that IS the id is not a name (t66): the row editor says
+                # "(no name yet)" instead of offering the id twice.
+                name = ""
             detected.setdefault(
                 package,
                 {
-                    "name": str(app.get("name") or "").strip(),
+                    "name": name,
                     "icon": str(app.get("icon") or "").strip(),
                 },
             )
diff --git a/custom_components/nvidia_shield/connection.py b/custom_components/nvidia_shield/connection.py
index c39c872..cc95097 100644
--- a/custom_components/nvidia_shield/connection.py
+++ b/custom_components/nvidia_shield/connection.py
@@ -39,11 +39,15 @@ from .const import (
     APP_ICON_FALLBACKS,
     APP_ICON_GENERIC,
     APP_LABEL_EXIT_MARKER,
+    APP_LABEL_GREP_SHAPES,
+    APP_LABEL_PROBE_LIMIT,
     APP_LABEL_TTL,
     APP_LIST_TTL,
     APP_SWEEP_CONCURRENCY,
     APP_SWEEP_SPACING,
+    CONF_APP_NAME_MAP,
     CONF_LOGCAT_EVENTS,
+    DEFAULT_APP_DIRECTORY,
     LOGCAT_SUPPRESS_WINDOW,
     STATE_STORE_TTL,
     SYSTEM_MEDIA_PACKAGES,
@@ -54,8 +58,9 @@ from .ha_adb_common import (
     icon_for_package,
     installed_packages,
     is_launchable,
-    parse_application_label,
+    parse_app_label,
     parse_audio_playback,
+    resolve_package_name,
 )
 
 _LOGGER = logging.getLogger(__name__)
@@ -224,6 +229,16 @@ class ShieldConnectionManager:
         # t55: packages whose label fetch is queued or in flight — a render
         # during the fetch must not enqueue the same package twice.
         self._app_label_inflight: set[str] = set()
+        # t66: the DEVICE-level "this firmware exposes no labels" memo. Set after
+        # APP_LABEL_PROBE_LIMIT consecutive probes that RAN and found nothing (one
+        # label-less package proves nothing — see _dump_label), and it expires
+        # with APP_LABEL_TTL so a changed device is re-checked within a day.
+        self._labels_absent_at: float | None = None
+        self._label_negative_streak: int = 0
+        # t66: packages whose probe FAILED (transport delivered nothing) and which
+        # therefore still owe a retry even while the memo above is set — t57's
+        # "a failed fetch is never cached as an answer" survives the memo.
+        self._app_label_failed: set[str] = set()
         # t57: the discovered installed-apps table — (rows, stamp), or None when
         # no sweep has produced a fresh answer. Only a sweep that RAN writes
         # here (t62 F1): a failed or unavailable sweep records itself as UNKNOWN
@@ -624,6 +639,68 @@ class ShieldConnectionManager:
             self._app_label_requests.add(package)
         return None
 
+    def _probe_still_owed(self, package: str) -> bool:
+        """True when a package's only remaining question is a RETRY (t66).
+
+        A probe that FAILED (no poller, shell error, no output at all) records
+        the package here and caches nothing, so the t57 UNIT 0 guarantee — a
+        failed fetch is retried, never stamped as an answer — survives the
+        device-level memo below. A probe that RAN and found no label is not owed
+        anything: it answered.
+        """
+        return package in self._app_label_failed
+
+    def _labels_absent(self) -> bool:
+        """The device-level "no labels here" memo, expiring with APP_LABEL_TTL.
+
+        A conclusion about a piece of firmware cannot be permanent — the user may
+        swap the device, update it, or add an app that does carry a label — so
+        after APP_LABEL_TTL the memo clears itself and one probe re-checks the
+        device (t66).
+        """
+        if self._labels_absent_at is None:
+            return False
+        if self._clock() - self._labels_absent_at > APP_LABEL_TTL:
+            self._labels_absent_at = None  # re-probe once per TTL
+            self._label_negative_streak = 0
+            return False
+        return True
+
+    @property
+    def app_labels_absent(self) -> bool:
+        """Public read of the memo (tests, and the options page's own note)."""
+        return self._labels_absent_at is not None and (
+            self._clock() - self._labels_absent_at <= APP_LABEL_TTL
+        )
+
+    def _entry_app_name_map(self) -> dict:
+        """The user's own apps map from the entry options ({} when unset)."""
+        options = getattr(self.entry, "options", None) or {}
+        return dict(options.get(CONF_APP_NAME_MAP) or {})
+
+    def _resolved_app_name(self, package: str, device_label=None, user_map=None) -> str:
+        """The display name for a discovered app, or "" when nothing names it (t66).
+
+        The ONE place the discovery table's name comes from, so it obeys the same
+        chain as the media player and the Current App sensor
+        (:func:`~.ha_adb_common.app_label.resolve_package_name`).
+
+        "" is deliberate: the row's ``package`` field already carries the id, and
+        a name that repeats it is exactly the ``pkg=pkg`` row the user reported
+        from the apps editor. Every consumer already reads an empty name as "this
+        app needs a name" (the editor says so out loud; the installed-apps sensor
+        publishes it as-is).
+        """
+        if user_map is None:
+            user_map = self._entry_app_name_map()
+        name = resolve_package_name(
+            package,
+            user_map=user_map,
+            device_label=device_label,
+            curated=DEFAULT_APP_DIRECTORY,
+        )
+        return "" if not name or name == package else name
+
     def _known_label(self, package: str) -> tuple[bool, str]:
         """``(known, label)`` from the label cache — "known" = a fetch RAN (t57).
 
@@ -682,28 +759,66 @@ class ShieldConnectionManager:
         except Exception as err:
             _LOGGER.debug("protocol device info unavailable: %s", err)
 
-    async def _dump_label(self, package: str) -> str | None:
-        """One `pm dump` label read: the label, or None for a FAILED fetch.
+    async def _dump_label(self, package: str, *, on_demand: bool = False) -> str | None:
+        """One label read: the label, or None for a FAILED fetch (t66).
 
-        The ONE place the dump is issued, shared by the on-demand path
+        The ONE place the probe is issued, shared by the on-demand path
         (:meth:`_fetch_app_label`) and the discovery sweep, so the t57 UNIT 0
         distinction — "the command ran and found no label" versus "the
         transport delivered nothing" — cannot drift between them. A successful
         fetch is also cached here, so both paths benefit from the 24 h cache and
         neither double-stamps.
 
-        Returns the label (possibly "") when the dump RAN, and None when it did
+        WHAT CHANGED IN t66, and why the old shape could not work on the user's
+        Shield: the probe used to ship ``pm dump <pkg> | head -300`` and parse the
+        answer with ``parse_application_label`` alone. A real dump on that device
+        is ~2241 lines of dumpsys, so ``head -300`` cut before the package
+        sections even appeared — the "no label" answer was a truncation
+        artefact, not a fact about the app. The probe now searches ON THE DEVICE
+        for every label shape Android prints (``APP_LABEL_GREP_SHAPES``) and only
+        those lines cross the wire, which is both smaller and honest.
+
+        A device that has already answered "no labels" for a run of packages is
+        NOT probed again (``APP_LABEL_PROBE_LIMIT``, expiring with
+        ``APP_LABEL_TTL``): the sweep's per-package fan-out stops there and a
+        package the device has never been asked about is answered "" from the
+        memo, with no shell at all. The two exceptions are deliberate: a package
+        whose probe FAILED still owes its retry (``_probe_still_owed``), and that
+        retry is only ever driven by the app ON SCREEN (``on_demand``) — one
+        shell, once, user-driven, then cached.
+        The two meanings of "" must stay distinct in the reader's head: a probe
+        that RAN and found nothing is a cached answer (zero further shells), and a
+        probe that COULD NOT run is not cached at all — it is recorded as owed and
+        retried for the app on screen.
+
+        Returns the label (possibly "") when the probe RAN, and None when it did
         not run or delivered no output at all.
 
         The exit marker is a liveness tell, not a status code (t62 nit b): it sits
-        after a ``head`` pipeline, and the Android shell that runs this has no
-        PIPESTATUS, so ``$?`` is ``head``'s status rather than the dump's. Nothing
-        reads its VALUE — ``parse_application_label`` strips the marker line — and
-        the only property tested here is that a command which RAN leaves non-blank
-        output even when the device prints no label. That non-blankness is what
+        after a pipeline, and the Android shell that runs this has no PIPESTATUS,
+        so ``$?`` is the last command's status rather than the dump's. Nothing
+        reads its VALUE — the parser ignores the marker line — and the only
+        property tested here is that a command which RAN leaves non-blank output
+        even when the device prints no label. That non-blankness is what
         separates "ran, no label" from "the transport delivered nothing", which
         this method turns into a cached negative versus a retry.
         """
+        if self._labels_absent():
+            if not self._probe_still_owed(package):
+                # The firmware already answered; a shell here would spend a round
+                # trip on a settled question (t66). The negative is cached so the
+                # render path stops asking cycle after cycle.
+                self.cache_app_label(package, "")
+                return ""
+            if not on_demand:
+                # a retry is not the sweep's job: the fan-out stops here
+                return ""
+            _LOGGER.debug(
+                "retrying the failed label probe for %s on the app now on "
+                "screen (the device reports no labels, but this one never "
+                "answered)",
+                package,
+            )
         poller = self._poller
         if poller is None or not self.connected:
             _LOGGER.debug(
@@ -711,12 +826,14 @@ class ShieldConnectionManager:
                 "retry on the next render",
                 package,
             )
+            self._app_label_failed.add(package)
             return None
         try:
             import shlex
 
             dump = await poller.shell(
-                f"pm dump {shlex.quote(package)} 2>/dev/null | head -300; "
+                f"pm dump {shlex.quote(package)} 2>/dev/null | "
+                f"grep -E {shlex.quote(APP_LABEL_GREP_SHAPES)} | head -40; "
                 f"echo {APP_LABEL_EXIT_MARKER}$?"
             )
         except Exception as err:
@@ -726,6 +843,7 @@ class ShieldConnectionManager:
                 package,
                 err,
             )
+            self._app_label_failed.add(package)
             return None
         if dump is None or not dump.strip():
             # the transport never delivered output: a FAILED fetch, not an
@@ -735,12 +853,37 @@ class ShieldConnectionManager:
                 "the next render",
                 package,
             )
+            self._app_label_failed.add(package)
             return None
-        label = parse_application_label(dump)
-        # F1: label-less pm dumps negatively cache too (empty label) so a
-        # later cycle performs zero additional pm dumps for the package.
+        self._app_label_failed.discard(package)  # it answered — nothing owed
+        label = parse_app_label(dump)
+        if label:
+            # This firmware DOES name apps: a run of misses before it proved nothing.
+            self._label_negative_streak = 0
+        else:
+            self._label_negative_streak += 1
+            if (
+                self._label_negative_streak >= APP_LABEL_PROBE_LIMIT
+                and self._labels_absent_at is None
+            ):
+                # A RUN of label-less answers, each of which really ran: this
+                # device keeps its labels somewhere we cannot read, so every
+                # further per-package probe is a shell spent on a settled
+                # question (t66). The conclusion expires with APP_LABEL_TTL.
+                self._labels_absent_at = self._clock()
+                _LOGGER.info(
+                    "%d label probes in a row found no app label — this device "
+                    "exposes none; names now come from the curated map and the "
+                    "user's own apps map for the next %s seconds",
+                    self._label_negative_streak,
+                    APP_LABEL_TTL,
+                )
+        # F1: label-less probes negatively cache too (empty label) so a later
+        # cycle performs zero additional probes for the package.
         self.cache_app_label(package, label)
-        return label
+        # "" (not None) is the "the probe RAN" answer, even when the device named
+        # nothing: None is reserved for a fetch that could not run (t57 UNIT 0).
+        return label or ""
 
     async def _fetch_app_label(self, package: str) -> None:
         """One on-demand `pm dump` label fetch; stamps ONLY a fetch that RAN.
@@ -752,7 +895,7 @@ class ShieldConnectionManager:
         """
         self._app_label_inflight.add(package)
         try:
-            await self._dump_label(package)
+            await self._dump_label(package, on_demand=True)
         finally:
             self._app_label_inflight.discard(package)
 
@@ -1049,10 +1192,16 @@ class ShieldConnectionManager:
         packages = [p for p in installed_packages(user_output, launcher_output)
                     if is_launchable(p)]
         labels = await self._sweep_labels(packages)
+        # t66: ONE precedence chain for the table too — the user's own map, then
+        # the device's label (when it exposes one), then the curated map, then the
+        # package id. An app nothing can name gets name "" rather than its own id:
+        # the row already carries the id in ``package``, and writing it twice is
+        # the ``pkg=pkg`` the user read as a bug in the apps editor.
+        user_map = self._entry_app_name_map()
         rows = [
             {
                 "package": package,
-                "name": labels.get(package) or package,
+                "name": self._resolved_app_name(package, labels.get(package), user_map),
                 "icon": icon_for_package(package, APP_ICON_FALLBACKS, APP_ICON_GENERIC),
             }
             for package in packages
diff --git a/custom_components/nvidia_shield/const.py b/custom_components/nvidia_shield/const.py
index efa2873..73ec0bd 100644
--- a/custom_components/nvidia_shield/const.py
+++ b/custom_components/nvidia_shield/const.py
@@ -3,7 +3,6 @@
 # Re-export shared Media DAG constants
 from .ha_adb_common.const import (  # noqa: F401
     ATTR_MEDIA_DAG_JELLYFIN_ENTITY,
-    ATTR_MEDIA_DAG_NOW_PLAYING,
     ATTR_MEDIA_DAG_RECEIVER_ENTITY,
     ATTR_MEDIA_DAG_RECEIVER_INPUT,
     ATTR_MEDIA_DAG_ROLE,
@@ -128,6 +127,26 @@ APP_LABEL_TTL = 24 * 3600
 # unlikely to collide with a manifest line.
 APP_LABEL_EXIT_MARKER = "__NSH_EXIT:"
 
+# t66: the label candidates searched FOR ON THE DEVICE. A real `pm dump` on the
+# user's Shield is ~2241 lines of dumpsys (package + activity + meminfo +
+# procstats + usagestats + batterystats), and the old fetch shipped
+# ``| head -300`` — which cut before the package sections even appeared, so the
+# negative was a truncation artefact rather than an answer. Searching on-device
+# keeps the wire small AND the answer honest: these are the three shapes Android
+# prints for an app's label (see ha_adb_common/app_label.py for what counts as a
+# name in each).
+APP_LABEL_GREP_SHAPES = "application-label|nonLocalizedLabel|labelRes"
+
+# t66: how many label probes may come back "ran, no label" IN A ROW before the
+# device is treated as one that exposes no labels at all (this user's Shield).
+# One label-less package is ordinary (system/OEM packages often have none), so a
+# single miss proves nothing — but a run of them means the whole firmware keeps
+# its labels in resources we cannot read, and every further ``pm dump`` per
+# package is a shell spent on a question already answered. The conclusion expires
+# with APP_LABEL_TTL so a device change is noticed within a day, and any label
+# found anywhere resets the run.
+APP_LABEL_PROBE_LIMIT = 5
+
 # P4-2a adb_command service: rawADB shell on this entry — OFF by default
 # (same security posture as the core androidtv integration).
 CONF_ADB_COMMAND_ENABLED = "adb_command_enabled"
@@ -142,13 +161,38 @@ CONF_LOGCAT_EVENTS = "logcat_events"
 # re-stamped (repeat suppression).
 LOGCAT_SUPPRESS_WINDOW = 15.0
 
-# P4-2c BrowseMedia app directory fallback when the options map is empty
-# (curated; the packages are the plan's reference apps).
+# P4-2c/t66: the curated package -> name table. It is BOTH the BrowseMedia/source
+# directory fallback and the third tier of the app-name chain (after the user's
+# own map and a device label, before the raw package id) — on a Shield whose
+# `pm dump` carries no label at all it is the answer that makes names work.
+#
+# It is a community-maintained convenience, NOT device truth: package ids are
+# stable, display names change with vendors and app updates, and an app this
+# table has never heard of falls back to its package id (shown once). Users
+# override any entry in Configure > Apps, and that edit always wins. Adding a row
+# is deliberately trivial — keep the ids sorted so the diff stays reviewable.
 DEFAULT_APP_DIRECTORY = {
     "ar.tvplayer.tv": "TiviMate",
+    "be.digitalia.mediasession2mqtt": "MediaSession2MQTT",
+    "com.amazon.amazonvideo.livingroom": "Prime Video",
+    "com.esaba.downloader": "Downloader",
     "com.github.damontecres.wholphin": "Wholphin",
-    "com.netflix.ninja": "Netflix",
+    "com.google.android.sambadocumentsprovider": "Samba (SMB)",
+    "com.google.android.speech.pumpkin": "Google Speech Services",
+    "com.google.android.tts": "Speech Services by Google",
+    "com.google.android.tvlauncher": "Google TV launcher",
     "com.google.android.youtube.tv": "YouTube",
+    "com.hyperion.grabber": "Hyperion Grabber",
+    "com.lonelycatgames.Xplore": "X-plore File Manager",
+    "com.nendo.argosy": "Argosy",
+    "com.netflix.ninja": "Netflix",
+    "com.plexapp.mediaserver.smb": "Plex Media Server",
+    "com.retroarch": "RetroArch",
+    "com.spocky.projengmenu": "Projectivy Launcher",
+    "dev.jdtech.jellyfin": "Findroid",
+    "dev.trooped.tvquickbars": "TV Quick Actions",
+    "flar2.homebutton": "Button Mapper",
+    "tv.pluto.android": "Pluto TV",
 }
 
 # t57: best-effort icon per discovered app. `pm dump` does NOT expose the
diff --git a/custom_components/nvidia_shield/ha_adb_common/__init__.py b/custom_components/nvidia_shield/ha_adb_common/__init__.py
index 409f371..724e0b6 100644
--- a/custom_components/nvidia_shield/ha_adb_common/__init__.py
+++ b/custom_components/nvidia_shield/ha_adb_common/__init__.py
@@ -1,10 +1,15 @@
 """Shared ADB utilities for Home Assistant custom integrations."""
 
 from .adb_client import AdbClientBase, parse_wakefulness
-from .app_label import parse_application_label
+from .app_label import (
+    parse_app_label,
+    parse_application_label,
+    parse_label_res,
+    parse_non_localized_label,
+    resolve_package_name,
+)
 from .const import (
     ATTR_MEDIA_DAG_JELLYFIN_ENTITY,
-    ATTR_MEDIA_DAG_NOW_PLAYING,
     ATTR_MEDIA_DAG_RECEIVER_ENTITY,
     ATTR_MEDIA_DAG_RECEIVER_INPUT,
     ATTR_MEDIA_DAG_ROLE,
@@ -31,7 +36,6 @@ from .running_apps import running_apps_from_sessions, running_apps_from_tasks
 
 __all__ = [
     "ATTR_MEDIA_DAG_JELLYFIN_ENTITY",
-    "ATTR_MEDIA_DAG_NOW_PLAYING",
     "ATTR_MEDIA_DAG_RECEIVER_ENTITY",
     "ATTR_MEDIA_DAG_RECEIVER_INPUT",
     "ATTR_MEDIA_DAG_ROLE",
@@ -41,7 +45,11 @@ __all__ = [
     "PlaybackIntent",
     "ensure_adb_key",
     "parse_all_media_sessions",
+    "parse_app_label",
     "parse_application_label",
+    "parse_label_res",
+    "parse_non_localized_label",
+    "resolve_package_name",
     "parse_audio_playback",
     "parse_foreground",
     "foreground_from_logcat",
diff --git a/custom_components/nvidia_shield/ha_adb_common/app_label.py b/custom_components/nvidia_shield/ha_adb_common/app_label.py
index d61fe56..d7f064c 100644
--- a/custom_components/nvidia_shield/ha_adb_common/app_label.py
+++ b/custom_components/nvidia_shield/ha_adb_common/app_label.py
@@ -1,19 +1,55 @@
-"""App friendly-name resolution from the package manifest (P5-1).
+"""App friendly-name resolution from the package dump (P5-1, t66).
 
-Parses `pm dump <package>` output for the ``application-label`` line — the
-default-locale variant; localized suffixed variants
-(``application-label-<lang>:``) are skipped, and multi-label blocks across
-locales are tolerated (first default-locale label wins), mirroring the
-tolerant pattern of the androidtv-adb core helper.
+Parses the label candidate LINES of a `pm dump <package>` / `dumpsys package`
+answer. Three shapes are searched, in priority order, because Android firmware
+prints different ones (and one of them on this user's Shield prints nothing at
+all):
+
+1. ``application-label: 'Netflix'`` — the AAPT-style manifest label, with its
+   localized suffixed variants (``application-label-<lang>:``) as a fallback.
+2. ``nonLocalizedLabel=Netflix`` — the ApplicationInfo field. It is usually
+   ``null`` on a TV device (the label comes from a resource), so a null/empty
+   value is not a name.
+3. ``labelRes=0x7f0f0001`` — a RESOURCE ID, not text. A dump can only name the
+   app here if it prints a quoted string, so a numeric/hex id is deliberately
+   NOT used as a name: we would be showing ``0x7f0f0001`` as if it were a name.
+
+Nothing here is a guess about which line exists: every branch is anchored on the
+literal prefix Android prints, and a miss returns None so the caller falls
+through to the curated map (the answer on a device with no labels at all).
+
+Also exports the ONE package -> name precedence chain, so the media player, the
+Current App sensor, the discovery table and the options editor cannot drift:
+user map > device label > curated map > raw package id.
 """
 
 import re
 
+#: ``application-label: 'X'`` / ``application-label-de: "X"`` — the AAPT form.
 _APPLICATION_LABEL_RE = re.compile(
     r"application-label(?:-[A-Za-z_-]+)?:\s*'([^']*)'"
     r'|application-label(?:-[A-Za-z_-]+)?:\s*"([^"]*)"'
 )
 
+#: ``nonLocalizedLabel=X`` — bare word, quoted string, or ``null``.
+_NON_LOCALIZED_LABEL_RE = re.compile(
+    r"nonLocalizedLabel=(?:'([^']*)'|\"([^\"]*)\"|([^\s,}]+))"
+)
+
+#: ``labelRes=0x7f0f0001`` — only a QUOTED value is text; an id is not a name.
+_LABEL_RES_RE = re.compile(r"labelRes=(?:'([^']*)'|\"([^\"]*)\")")
+
+#: Values a dump prints that are not names.
+_NOT_A_NAME = ("", "null", "0x0", "0")
+
+
+def _clean(value: str | None) -> str | None:
+    """A candidate label, or None when it is empty/``null``."""
+    text = (value or "").strip()
+    if text.lower() in _NOT_A_NAME:
+        return None
+    return text
+
 
 def parse_application_label(dump: str) -> str | None:
     """The package's default-locale application label, or None.
@@ -24,7 +60,92 @@ def parse_application_label(dump: str) -> str | None:
     if not dump:
         return None
     for match in _APPLICATION_LABEL_RE.finditer(dump):
-        label = (match.group(1) or match.group(2) or "").strip()
-        if label and label not in ("null", ""):
+        label = _clean(match.group(1) or match.group(2))
+        if label:
             return label
     return None
+
+
+def parse_non_localized_label(dump: str) -> str | None:
+    """The ``nonLocalizedLabel=`` value, or None (t66).
+
+    ``null`` is the common case on a TV and is NOT a name.
+    """
+    if not dump:
+        return None
+    for match in _NON_LOCALIZED_LABEL_RE.finditer(dump):
+        label = _clean(match.group(1) or match.group(2) or match.group(3))
+        if label:
+            return label
+    return None
+
+
+def parse_label_res(dump: str) -> str | None:
+    """A QUOTED ``labelRes=`` value, or None (t66).
+
+    ``labelRes=0x7f0f0001`` is a resource id — showing it as an app name would be
+    worse than showing the package id, so only a quoted string counts.
+    """
+    if not dump:
+        return None
+    for match in _LABEL_RES_RE.finditer(dump):
+        label = _clean(match.group(1) or match.group(2))
+        if label:
+            return label
+    return None
+
+
+def parse_app_label(dump: str) -> str | None:
+    """The best device label in ``dump``, trying every shape Android prints.
+
+    Priority follows how specific the shape is: the AAPT manifest label first,
+    then the ApplicationInfo field, then a quoted resource label. None means the
+    device did not name this app — which, on a device whose dump carries no
+    labels at all, is the definitive answer that sends the caller to the curated
+    map (t66).
+    """
+    return (
+        parse_application_label(dump)
+        or parse_non_localized_label(dump)
+        or parse_label_res(dump)
+    )
+
+
+def resolve_package_name(
+    package: str | None,
+    *,
+    user_map=None,
+    device_label=None,
+    curated=None,
+) -> str | None:
+    """THE package -> display-name chain (t66): one order, every surface.
+
+    1. the user's explicit apps map (their own edit always wins);
+    2. the device's own label, when the device exposes one;
+    3. the curated community map (the answer on a device with no labels: see
+       ``DEFAULT_APP_DIRECTORY``);
+    4. the raw package id, shown ONCE.
+
+    ``device_label`` may be a VALUE or a CALLABLE. A callable is looked up only
+    after the user map has missed, which is what keeps tier 2 LAZY: the manager's
+    ``app_label()`` accessor enqueues an on-demand device fetch on a miss, so
+    calling it for an app the user has already named would spend a shell to learn
+    something nobody reads.
+
+    Returns None only for an empty package, so a caller can always display
+    something. The id is never doubled into a ``pkg=pkg`` row: tier 4 IS the id,
+    and there is no tier that formats it as a name as well.
+    """
+    if not package:
+        return None
+    mapped = (user_map or {}).get(package)
+    if mapped:
+        return str(mapped)
+    raw_label = device_label(package) if callable(device_label) else device_label
+    label = _clean(raw_label)
+    if label:
+        return label
+    curated_name = (curated or {}).get(package)
+    if curated_name:
+        return str(curated_name)
+    return package
diff --git a/custom_components/nvidia_shield/ha_adb_common/const.py b/custom_components/nvidia_shield/ha_adb_common/const.py
index 79df930..938e659 100644
--- a/custom_components/nvidia_shield/ha_adb_common/const.py
+++ b/custom_components/nvidia_shield/ha_adb_common/const.py
@@ -17,9 +17,11 @@ ATTR_MEDIA_DAG_RECEIVER_INPUT = "media_dag_receiver_input"
 # Jellyfin media_player entity for rich metadata
 ATTR_MEDIA_DAG_JELLYFIN_ENTITY = "media_dag_jellyfin_entity"
 
-# Now-playing dict: {state, title, media_type, series?, season?, episode?,
-#                    duration?, position?, thumbnail?, artist?, album?, media_info?, source?}
-ATTR_MEDIA_DAG_NOW_PLAYING = "media_dag_now_playing"
+# t65 flattened the now-playing reading into separate top-level ``now_playing_*``
+# attributes on the media player (``now_playing_via`` names the tier), so the single
+# nested ``media_dag_now_playing`` key this file used to define is GONE — the last
+# reference to it was removed in t66's cleanup. The flattened names live in
+# ``custom_components/nvidia_shield/media_player.py`` (``now_playing_attribute_names``).
 
 # List of running app dicts: [{package, state, title?, artist?}, ...]
 ATTR_MEDIA_DAG_RUNNING_APPS = "media_dag_running_apps"
diff --git a/custom_components/nvidia_shield/media_player.py b/custom_components/nvidia_shield/media_player.py
index dd57442..9539f72 100644
--- a/custom_components/nvidia_shield/media_player.py
+++ b/custom_components/nvidia_shield/media_player.py
@@ -116,7 +116,11 @@ from .const import (
     TIVIMATE_PACKAGE,
     command_scripts,
 )
-from .ha_adb_common import PlaybackIntent, running_apps_from_tasks
+from .ha_adb_common import (
+    PlaybackIntent,
+    resolve_package_name,
+    running_apps_from_tasks,
+)
 from .jellyfin import fetch_session_media_info
 from .power import discover_power_siblings
 
@@ -377,26 +381,29 @@ def resolve_app_name(
     package: str | None,
     app_name_map,
     app_label,
+    curated=None,
 ) -> str | None:
-    """Resolve a package id to a friendly name (t40; entity tier removed t41).
+    """Resolve a package id to a friendly name (t40; curated tier t66).
 
     The single chain behind ``media_player.app_name`` and the Current App
-    sensor:
+    sensor, delegating to ``ha_adb_common.app_label.resolve_package_name`` so the
+    discovery table and the options editor cannot drift from it:
 
     1. the user's explicit apps map (``CONF_APP_NAME_MAP``);
-    2. the cached device-manifest label (``manager.app_label`` — a cache miss
-       enqueues the on-demand fetch and falls through);
-    3. the raw package text — a bare or unknown package falls back to itself,
-       never to None.
+    2. the cached device label (``manager.app_label`` — a cache miss enqueues the
+       on-demand fetch and falls through);
+    3. the curated table (``DEFAULT_APP_DIRECTORY``) — the tier that names apps on
+       a Shield whose ``pm dump`` exposes no label at all;
+    4. the raw package text — a bare or unknown package falls back to itself,
+       never to None, and it is shown ONCE (no ``pkg=pkg`` anywhere).
     """
-    if package:
-        mapped = (app_name_map or {}).get(package)
-        if mapped:
-            return mapped
-        label = app_label(package) if app_label is not None else None
-        if label:
-            return label
-    return package or None
+    return resolve_package_name(
+        package,
+        user_map=app_name_map,
+        # the callable keeps the device tier LAZY: the user's map answers first
+        device_label=app_label,
+        curated=DEFAULT_APP_DIRECTORY if curated is None else curated,
+    )
 
 
 class NvidiaShieldMediaPlayer(MediaPlayerEntity):
diff --git a/custom_components/nvidia_shield/sensor.py b/custom_components/nvidia_shield/sensor.py
index bbb734b..727a889 100644
--- a/custom_components/nvidia_shield/sensor.py
+++ b/custom_components/nvidia_shield/sensor.py
@@ -124,9 +124,10 @@ class NvidiaShieldCurrentAppSensor(NvidiaShieldSensorBase):
         """The resolved friendly name (t40), not the raw package id.
 
         Exactly the ``media_player.app_name`` chain via the shared
-        ``resolve_app_name``: apps map → cached manifest label → the package
-        text itself. A bare or unknown package shows its package text rather
-        than None, like the media player's app_name.
+        ``resolve_app_name``: apps map → cached device label → the built-in app
+        names (t66) → the package text itself. A bare or unknown package shows its
+        package text rather than None, like the media player's app_name, and that
+        text is the id ONCE — never a ``pkg=pkg`` pair.
 
         This is why ``_package`` must be the CURRENT app and the label fetch
         must be driven off it (t55): the device-manifest label is resolved on
diff --git a/docs/RESEARCH_SHIELD.md b/docs/RESEARCH_SHIELD.md
index a9ba903..f34157f 100644
--- a/docs/RESEARCH_SHIELD.md
+++ b/docs/RESEARCH_SHIELD.md
@@ -404,7 +404,7 @@ covers what the external pieces provided:
 
 | External piece | What it provided | Now native (file) |
 |---|---|---|
-| `mediasession2mqtt` bridge sensors | per-app now-playing pushed over MQTT from the Shield's notification listener | ADB media-session polling + the foreground tier (`connection.py`, `media_player.py`); the `media_dag_now_playing` attribute carries `source: jellyfin \| media_session \| app` |
+| `mediasession2mqtt` bridge sensors | per-app now-playing pushed over MQTT from the Shield's notification listener | ADB media-session polling + the foreground tier (`connection.py`, `media_player.py`); the reading is published as separate top-level `now_playing_*` attributes, and `now_playing_via` names the tier (`jellyfin` \| `media_session` \| `app`) |
 | Universal media player `media_dag_running_apps` template | app switcher list with playback states | `running_apps_from_tasks(..., last_known=...)` (`ha_adb_common/running_apps.py`) — enriched with `foreground`/`state`/`last_known_state`/`last_known_title` |
 | Universal player "currently in app X" template | fallback state for no-session apps | the foreground ladder tier (`media_player.py` state property, tier 3.5) |
 | External androidtv_remote entity for commands | remote keys | the protocol-backed `remote` entity (Phase 2) — the external entity remains only as the transitional last command tier, removed in Phase 4 |
diff --git a/tests/test_command_routing.py b/tests/test_command_routing.py
index 7abdfb4..7afeb6a 100644
--- a/tests/test_command_routing.py
+++ b/tests/test_command_routing.py
@@ -875,7 +875,9 @@ class TestPrivacyOrdering:
         player._last_channel = "101"
         np = player._get_app_now_playing()
 
-        assert np["title"] == TIVIMATE_PACKAGE  # app identity, not the channel
+        # t66: the identity tier now resolves through the shared chain, so the app
+        # reads by its name — still the APP, never the channel or the media title.
+        assert np["title"] == "TiviMate"
         assert np["title"] != "101"
 
     async def test_channel_number_never_leaks_in_privacy_mode_with_last_known(self, modules):
@@ -893,7 +895,8 @@ class TestPrivacyOrdering:
         np = player._get_app_now_playing()
 
         assert np["state"] == "playing"  # playback truth survives privacy
-        assert np["title"] == TIVIMATE_PACKAGE  # but identity stays the app
+        assert np["title"] == "TiviMate"  # but identity stays the app
+        assert "Sky News" not in np.values()  # the media title never leaks
 
 
 class TestVolumeAndPowerLadder:
@@ -1725,23 +1728,26 @@ class TestCurrentAppNameEndToEndT55:
         return player, manager, sensor, poller
 
     async def test_known_app_resolves_by_fetching_its_label(self, modules):
+        # t66: a package NOT in the curated table, so the fall-through under test
+        # is really the raw id and not a curated name.
         player, manager, sensor, poller = self._wire(
-            modules, "com.netflix.ninja", {"com.netflix.ninja": "application-label: 'Netflix'"}
+            modules,
+            "com.example.labelled",
+            {"com.example.labelled": "application-label: 'Labelled App'"},
         )
 
         # first render: cache miss → the sensor falls through to the package
-        assert sensor.native_value == "com.netflix.ninja"
-        assert manager.app_label_requests == {"com.netflix.ninja"}
+        assert sensor.native_value == "com.example.labelled"
+        assert manager.app_label_requests == {"com.example.labelled"}
 
         # the watchdog drains the fetch for the CURRENT package
         await manager._watchdog_cycle(5.0)
         await asyncio.sleep(0)
-        assert poller.dumps and "com.netflix.ninja" in poller.dumps[0]
+        assert poller.dumps and "com.example.labelled" in poller.dumps[0]
 
         # next render: the resolved device name, in the sensor AND the player
-        assert sensor.native_value == "Netflix"
-        assert sensor.extra_state_attributes["app_name"] == "Netflix"
-        assert player.app_name == "Netflix"
+        assert sensor.native_value == "Labelled App"
+        assert player.app_name == "Labelled App"
 
     async def test_mapped_app_never_needs_the_device_label(self, modules):
         player, manager, sensor, poller = self._wire(
@@ -1818,7 +1824,9 @@ class TestCurrentAppNameEndToEndT55:
         here — the pre-fix tree fails this at the second cycle.
         """
         player, manager, sensor, poller = self._wire(
-            modules, "ar.tvplayer.tv", {"ar.tvplayer.tv": "application-label: 'TiviMate'"}
+            modules,
+            "com.example.labelled",
+            {"com.example.labelled": "application-label: 'Labelled App'"},
         )
 
         async def _blank(_cmd):
@@ -1827,16 +1835,16 @@ class TestCurrentAppNameEndToEndT55:
 
         # cycle 1: the transport delivered nothing → a FAILED fetch
         poller.shell = _blank
-        assert sensor.native_value == "ar.tvplayer.tv"
+        assert sensor.native_value == "com.example.labelled"
         await manager._watchdog_cycle(5.0)
         await asyncio.sleep(0)
 
-        assert "ar.tvplayer.tv" not in manager._app_labels  # nothing stamped
-        assert sensor.native_value == "ar.tvplayer.tv"
+        assert "com.example.labelled" not in manager._app_labels  # nothing stamped
+        assert sensor.native_value == "com.example.labelled"
 
         # the next render re-enqueues, and the retry is what recovers the label
-        assert manager.app_label("ar.tvplayer.tv") is None
-        assert manager.app_label_requests == {"ar.tvplayer.tv"}
+        assert manager.app_label("com.example.labelled") is None
+        assert manager.app_label_requests == {"com.example.labelled"}
         dump_count = len([d for d in poller.dumps if "pm dump" in d])
 
         real_shell = _LabelPoller.shell.__get__(poller, _LabelPoller)
@@ -1846,9 +1854,9 @@ class TestCurrentAppNameEndToEndT55:
 
         # a later GOOD dump recovers it — the pre-fix cache made this impossible
         assert len([d for d in poller.dumps if "pm dump" in d]) == dump_count + 1
-        assert sensor.native_value == "TiviMate"
-        assert sensor.extra_state_attributes["app_name"] == "TiviMate"
-        assert player.app_name == "TiviMate"
+        assert sensor.native_value == "Labelled App"
+        assert sensor.extra_state_attributes["app_name"] == "Labelled App"
+        assert player.app_name == "Labelled App"
 
     @pytest.mark.parametrize("blank", [None, ""], ids=["none", "empty"])
     async def test_the_player_recovers_the_label_after_a_failed_fetch(
@@ -1861,7 +1869,9 @@ class TestCurrentAppNameEndToEndT55:
         stuck on the package id — so both are pinned.
         """
         player, manager, sensor, poller = self._wire(
-            modules, "com.netflix.ninja", {"com.netflix.ninja": "application-label: 'Netflix'"}
+            modules,
+            "com.example.labelled",
+            {"com.example.labelled": "application-label: 'Labelled App'"},
         )
 
         async def _blank(_cmd):
@@ -1869,19 +1879,19 @@ class TestCurrentAppNameEndToEndT55:
             return blank
 
         poller.shell = _blank
-        assert player.app_name == "com.netflix.ninja"
+        assert player.app_name == "com.example.labelled"
         await manager._watchdog_cycle(5.0)
         await asyncio.sleep(0)
-        assert "com.netflix.ninja" not in manager._app_labels
-        assert player.app_name == "com.netflix.ninja"
+        assert "com.example.labelled" not in manager._app_labels
+        assert player.app_name == "com.example.labelled"
 
-        assert manager.app_label("com.netflix.ninja") is None  # re-enqueued
+        assert manager.app_label("com.example.labelled") is None  # re-enqueued
         poller.shell = _LabelPoller.shell.__get__(poller, _LabelPoller)
         await manager._watchdog_cycle(5.0)
         await asyncio.sleep(0)
 
-        assert player.app_name == "Netflix"
-        assert sensor.native_value == "Netflix"
+        assert player.app_name == "Labelled App"
+        assert sensor.native_value == "Labelled App"
 
     async def test_a_genuinely_label_less_package_is_still_negatively_cached(
         self, modules
@@ -1910,16 +1920,16 @@ class TestCurrentAppNameEndToEndT55:
     async def test_the_fetch_targets_the_app_on_screen_not_the_previous_one(self, modules):
         player, manager, sensor, poller = self._wire(
             modules,
-            "ar.tvplayer.tv",
+            "com.example.one",
             {
-                "ar.tvplayer.tv": "application-label: 'TiviMate'",
-                "com.google.android.youtube.tv": "application-label: 'YouTube'",
+                "com.example.one": "application-label: 'One TV'",
+                "com.example.two": "application-label: 'Two TV'",
             },
         )
 
-        assert sensor.native_value == "ar.tvplayer.tv"  # miss for the 1st app
-        manager._on_current_app("com.google.android.youtube.tv")
-        assert sensor.native_value == "com.google.android.youtube.tv"
+        assert sensor.native_value == "com.example.one"  # miss for the 1st app
+        manager._on_current_app("com.example.two")
+        assert sensor.native_value == "com.example.two"
 
         # both packages are queued; the watchdog drains ONE per cycle
         await manager._watchdog_cycle(5.0)
@@ -1929,9 +1939,9 @@ class TestCurrentAppNameEndToEndT55:
 
         # every drained fetch names a package, and the app that is CURRENT is
         # the one the sensor now shows the label of
-        assert "ar.tvplayer.tv" in poller.dumps[0]
-        assert "com.google.android.youtube.tv" in poller.dumps[1]
-        assert sensor.native_value == "YouTube"
+        assert "com.example.one" in poller.dumps[0]
+        assert "com.example.two" in poller.dumps[1]
+        assert sensor.native_value == "Two TV"
 
 
 class TestAvailabilityLadder:
diff --git a/tests/test_config_flow_ux.py b/tests/test_config_flow_ux.py
index 96579a5..d8c7fba 100644
--- a/tests/test_config_flow_ux.py
+++ b/tests/test_config_flow_ux.py
@@ -2250,8 +2250,9 @@ class TestDiscoveryFeedsTheAppsTable:
             [
                 {"package": "ar.tvplayer.tv", "name": "TiviMate", "icon": "mdi:television-classic"},
                 {"package": "org.xbmc.kodi", "name": "Kodi", "icon": "mdi:kodi"},
-                {"package": "com.example.handmade", "name": "com.example.handmade",
-                 "icon": "mdi:television"},
+                # t66: a discovered app nothing can name comes through as name ""
+                # (the sweep no longer repeats the package id as a name)
+                {"package": "com.example.handmade", "name": "", "icon": "mdi:television"},
             ],
         )
 
@@ -2263,7 +2264,9 @@ class TestDiscoveryFeedsTheAppsTable:
         assert lines[0] == "ar.tvplayer.tv=Telly"
         # every discovered app is a row too — this is the live bug
         assert "org.xbmc.kodi=Kodi" in lines
-        assert "com.example.handmade=com.example.handmade" in lines
+        # ...but an app nothing names is NOT pre-filled as `pkg=pkg` (t66): the id
+        # would be stored as if it were a name. The row editor still offers it.
+        assert not [line for line in lines if line.startswith("com.example.handmade=")]
         # the user's name wins over the device label for the same package
         assert not [line for line in lines if line == "ar.tvplayer.tv=TiviMate"]
         # ...and the discovered icon is NEVER written into the icon column
diff --git a/tests/test_connection.py b/tests/test_connection.py
index 8856873..4fdaa8f 100644
--- a/tests/test_connection.py
+++ b/tests/test_connection.py
@@ -18,7 +18,7 @@ from nsh.connection import (
     ShieldConnectionManager,
     StateStore,
 )
-from nsh.const import ADB_POLL_INTERVAL, APP_LABEL_EXIT_MARKER
+from nsh.const import ADB_POLL_INTERVAL, APP_LABEL_EXIT_MARKER, APP_LABEL_GREP_SHAPES
 
 
 def test_connection_module_depends_on_no_homeassistant():
@@ -310,12 +310,16 @@ class TestAppLabelCache:
 
         await asyncio.sleep(0)  # drain the fetch task
 
-        # t57: the dump carries the exit marker too, so a blank answer can be
-        # told from a FAILED fetch; the package is still quoted
+        # t57: the probe carries the exit marker too, so a blank answer can be told
+        # from a FAILED fetch; the package is still quoted. t66: the label candidates
+        # are filtered ON THE DEVICE, and the 2241-line dump is never truncated —
+        # `head -300` cut it before the package section on the user's Shield.
         assert manager._poller.shell_calls == [
-            f"pm dump {shlex.quote(wicked)} 2>/dev/null | head -300; "
+            f"pm dump {shlex.quote(wicked)} 2>/dev/null | "
+            f"grep -E {shlex.quote(APP_LABEL_GREP_SHAPES)} | head -40; "
             f"echo {APP_LABEL_EXIT_MARKER}$?"
         ]
+        assert "head -300" not in manager._poller.shell_calls[0]
 
 
 class TestLogcatStream:
diff --git a/tests/test_installed_apps.py b/tests/test_installed_apps.py
index efbfc3f..b910336 100644
--- a/tests/test_installed_apps.py
+++ b/tests/test_installed_apps.py
@@ -37,6 +37,8 @@ def scoped_entity_stubs():
 from nsh.connection import ShieldConnectionManager  # noqa: E402
 from nsh.const import (  # noqa: E402
     APP_ICON_FALLBACKS,
+    APP_LABEL_EXIT_MARKER,
+    APP_LABEL_GREP_SHAPES,
     APP_LIST_TTL,
     CONF_APP_NAME_MAP,
 )
@@ -50,16 +52,35 @@ from nsh.ha_adb_common import (  # noqa: E402
 
 # ── realistic device output ────────────────────────────────────────────────
 
-#: `pm list packages -3` on a Shield with a normal app set. Wrapped the way a
-#: real adb shell answers (a bare list, no header).
+#: `pm list packages -3` on the user's Shield (t66). These are the apps the
+#: report listed, which is also why they are the curated table's rows: on THIS
+#: device the firmware names none of them, so the table — not the device — is where
+#: the names come from. `org.xbmc.kodi` and `com.example.handmade` are kept as the
+#: packages NOTHING can name (no device label, no curated row), so the fall-through
+#: to "" is pinned on the same list.
 USER_PACKAGES = (
     "package:ar.tvplayer.tv\n"
+    "package:be.digitalia.mediasession2mqtt\n"
+    "package:com.amazon.amazonvideo.livingroom\n"
+    "package:com.esaba.downloader\n"
     "package:com.github.damontecres.wholphin\n"
-    "package:com.netflix.ninja\n"
+    "package:com.google.android.sambadocumentsprovider\n"
+    "package:com.google.android.speech.pumpkin\n"
+    "package:com.google.android.tts\n"
+    "package:com.google.android.tvlauncher\n"
     "package:com.google.android.youtube.tv\n"
+    "package:com.hyperion.grabber\n"
+    "package:com.lonelycatgames.Xplore\n"
+    "package:com.nendo.argosy\n"
+    "package:com.netflix.ninja\n"
+    "package:com.plexapp.mediaserver.smb\n"
+    "package:com.retroarch\n"
+    "package:com.spocky.projengmenu\n"
+    "package:dev.jdtech.jellyfin\n"
+    "package:dev.trooped.tvquickbars\n"
+    "package:flar2.homebutton\n"
+    "package:tv.pluto.android\n"
     "package:org.xbmc.kodi\n"
-    "package:com.plexapp.android\n"
-    "package:com.spotify.tv.android\n"
     "package:com.example.handmade\n"
 )
 
@@ -88,18 +109,133 @@ DUMP_LABEL_LESS = (
     "  codePath=/data/app/com.example.handmade\n"
 )
 
+#: The REAL device shape, trimmed: `pm dump com.netflix.ninja` on the user's Shield
+#: (captured 2026-10-12; the observed structure is recorded in
+#: ``docs/INTENTS_PLAN.md`` §1). It is the evidence behind this work — the raw
+#: ~2241-line capture text is NOT in the repository, so this is a faithful
+#: reproduction of the recorded structure, not the capture itself.
+#:
+#: WHAT IS REPRODUCED: the whole-dumpsys envelope (`DUMP OF SERVICE package:`,
+#: then `activity`, `meminfo`, `procstats`, `usagestats`, `batterystats`), the
+#: package section opening with `Activity Resolver Table:` followed by the
+#: receiver/service tables, Netflix's real entries (`cast:`, `http`/`https` with
+#: `www.netflix.com` path globs, `nflx:`, `netflix:`, `MAIN` +
+#: `LEANBACK_LAUNCHER`) and the capture's own noise (`com.netflix.action.DEBUG_*`,
+#: `androidx.*`, `com.google.firebase.*`, `ConstraintProxy*`) — and, the point of
+#: the whole exercise, NO `application-label` line anywhere: grepping the capture
+#: for `application-label` found nothing.
+#:
+#: (`nonLocalizedLabel=null`, `labelRes=0x7f0f0d31`) — the capture's record covers
+#: the resolver tables, not this block. They are here because they are exactly the
+#: shapes the on-device search must NOT read as a name: `null` is not a name, and a
+#: resource id is not a name either. The bodies of the four trailing service
+#: sections are trimmed to one line each (the capture is ~2241 lines; this is ~80).
+DUMP_NETFLIX_REAL = (
+    "DUMP OF SERVICE package:\n"
+    "Activity Resolver Table:\n"
+    "  Full MIME Types:\n"
+    "    video/*: 7f0f4c12 com.netflix.ninja/.ui.MainActivity filter 7f0f0a11\n"
+    "      Action: \"android.intent.action.VIEW\"\n"
+    "      Category: \"android.intent.category.DEFAULT\"\n"
+    "      Category: \"android.intent.category.BROWSABLE\"\n"
+    "      Scheme: \"https\"\n"
+    "        Authority: \"www.netflix.com\": -1\n"
+    "        Path: \"PatternMatcher{LITERAL,/title}\"\n"
+    "      Scheme: \"http\"\n"
+    "        Authority: \"www.netflix.com\": -1\n"
+    "    nflx: 7f0f4c12 com.netflix.ninja/.ui.MainActivity filter 7f0f0a12\n"
+    "      Action: \"android.intent.action.VIEW\"\n"
+    "      Category: \"android.intent.category.DEFAULT\"\n"
+    "      Scheme: \"nflx\"\n"
+    "    netflix: 7f0f4c12 com.netflix.ninja/.ui.MainActivity filter 7f0f0a13\n"
+    "      Action: \"android.intent.action.VIEW\"\n"
+    "      Scheme: \"netflix\"\n"
+    "  Non-Data Actions:\n"
+    "    android.intent.action.MAIN:\n"
+    "      7f0f4c12 com.netflix.ninja/.ui.MainActivity filter 7f0f0a14\n"
+    "        Action: \"android.intent.action.MAIN\"\n"
+    "        Category: \"android.intent.category.LEANBACK_LAUNCHER\"\n"
+    "    cast:\n"
+    "      7f0f4c12 com.netflix.ninja/.ui.MainActivity filter 7f0f0a15\n"
+    "        Action: \"android.intent.action.VIEW\"\n"
+    "        Scheme: \"cast\"\n"
+    "    com.netflix.action.DEBUG_FORCE_EXIT:\n"
+    "      7f0f4c12 com.netflix.ninja/.DebugReceiver filter 7f0f0a16\n"
+    "        Action: \"com.netflix.action.DEBUG_FORCE_EXIT\"\n"
+    "Receiver Resolver Table:\n"
+    "  Non-Data Actions:\n"
+    "    androidx.work.DiagnosticsReportReceiver:\n"
+    "      7f0f4c12 com.netflix.ninja/androidx.work.impl.diagnostics.DiagnosticsReceiver\n"
+    "    com.google.firebase.MESSAGING_EVENT:\n"
+    "      7f0f4c12 com.netflix.ninja/com.google.firebase.messaging.FirebaseMessagingService\n"
+    "    com.netflix.action.DEBUG_PUSH_NOTIFICATION:\n"
+    "      7f0f4c12 com.netflix.ninja/.push.DebugPushReceiver\n"
+    "Service Resolver Table:\n"
+    "  Non-Data Actions:\n"
+    "    com.google.firebase.INSTANCE_ID_EVENT:\n"
+    "      7f0f4c12 com.netflix.ninja/com.google.firebase.iid.FirebaseInstanceIdService\n"
+    "Provider Resolver Table:\n"
+    "  Authorities:\n"
+    "    com.netflix.ninja.androidx-startup:\n"
+    "      7f0f4c12 com.netflix.ninja/androidx.startup.InitializationProvider\n"
+    "Permissions:\n"
+    "  Permission [com.netflix.ninja.C2D_MESSAGE]:\n"
+    "    sourcePackage=com.netflix.ninja\n"
+    "Packages:\n"
+    "  Package [com.netflix.ninja] (7f1a2b3c):\n"
+    "    userId=10123\n"
+    "    pkg=Package{7f1a2b3c com.netflix.ninja}\n"
+    # the two INFERRED lines (see the note above): the shapes the search must skip
+    "    nonLocalizedLabel=null\n"
+    "    labelRes=0x7f0f0d31\n"
+    "    versionCode=61102 minSdk=21 targetSdk=29\n"
+    "    versionName=6.11.2 build 3413\n"
+    "    codePath=/data/app/com.netflix.ninja-1/base.apk\n"
+    "    ConstraintProxy.BatteryBroadcastReceiver=\n"
+    "      com.netflix.ninja/androidx.work.impl.constraints.trackers."
+    "ConstraintProxy$BatteryChargingProxy\n"
+    "  Activity Resolver Table: (nothing else listed for this package)\n"
+    "DUMP OF SERVICE activity:\n"
+    "  ACTIVITY MANAGER PENDING INTENTS (dumpsys activity pending)\n"
+    "    (nothing)\n"
+    "DUMP OF SERVICE meminfo:\n"
+    "  Applications Memory Usage (in Kilobytes):\n"
+    "DUMP OF SERVICE procstats:\n"
+    "  Process Stats:\n"
+    "DUMP OF SERVICE usagestats:\n"
+    "  In-memory rollup:\n"
+    "DUMP OF SERVICE batterystats:\n"
+    "  Battery Stats:\n"
+)
+
+
+def _on_device_grep(dump: str) -> str:
+    """What the WIRE carries once the fetch greps on the device (t66).
+
+    The real command is ``pm dump <pkg> | grep -E <shapes> | head -40``, so the
+    2241-line dump never crosses ADB: only the matching lines do, plus the exit
+    marker the manager relies on. A fake that returned the whole dump would only
+    prove that the PARSER tolerates it — not that the manager never had to.
+    """
+    shapes = APP_LABEL_GREP_SHAPES.split("|")
+    hits = [line for line in dump.splitlines() if any(s in line for s in shapes)]
+    return "\n".join(hits[:40]) + f"\n{APP_LABEL_EXIT_MARKER}0\n"
+
 
 class _DiscoveryPoller:
     """A poller whose shell answers the discovery commands from fixtures."""
 
     def __init__(self, *, packages=USER_PACKAGES, launcher=LAUNCHER_RESOLVED,
-                 dumps=None, failure=False):
+                 dumps=None, failure=False, on_device_grep=False):
         self.connected = True
         self.connected_override = True
         self.packages = packages
         self.launcher = launcher
         self.dumps = dict(dumps or {})
         self.failure = failure
+        #: t66: True = the fixture values are RAW device dumps and the fake applies
+        #: the on-device grep before answering, like the real command does.
+        self.on_device_grep = on_device_grep
         self.shell_calls: list[str] = []
 
     @property
@@ -119,7 +255,7 @@ class _DiscoveryPoller:
 
             for package, dump in self.dumps.items():
                 if shlex.quote(package) in cmd or package in cmd:
-                    return dump
+                    return _on_device_grep(dump) if self.on_device_grep else dump
             return "__NSH_EXIT:0\n"  # ran, no label
         return None
 
@@ -132,15 +268,9 @@ class TestPackageListParsing:
     """`pm list packages -3`."""
 
     def test_real_output_parses_one_package_per_line(self):
+        # the order of `pm list packages -3` is preserved, one id per line
         assert parse_package_list(USER_PACKAGES) == [
-            "ar.tvplayer.tv",
-            "com.github.damontecres.wholphin",
-            "com.netflix.ninja",
-            "com.google.android.youtube.tv",
-            "org.xbmc.kodi",
-            "com.plexapp.android",
-            "com.spotify.tv.android",
-            "com.example.handmade",
+            line.removeprefix("package:") for line in USER_PACKAGES.splitlines()
         ]
 
     def test_noise_and_duplicates_are_tolerated(self):
@@ -192,10 +322,19 @@ class TestInstalledPackagesMerge:
     """User apps plus the active launcher, de-duplicated."""
 
     def test_launcher_is_added_last(self):
-        packages = installed_packages(USER_PACKAGES, LAUNCHER_RESOLVED)
+        # t66: the user's own list contains the launcher, so the append-last
+        # property is pinned on a list that does NOT already have it
+        user_only = USER_PACKAGES.replace("package:com.google.android.tvlauncher\n", "")
+        packages = installed_packages(user_only, LAUNCHER_RESOLVED)
         assert packages[-1] == "com.google.android.tvlauncher"
+        assert packages.count("com.google.android.tvlauncher") == 1
         assert packages[0] == "ar.tvplayer.tv"
 
+    def test_a_launcher_already_listed_keeps_its_place(self):
+        packages = installed_packages(USER_PACKAGES, LAUNCHER_RESOLVED)
+        assert packages.count("com.google.android.tvlauncher") == 1
+        assert packages == parse_package_list(USER_PACKAGES)
+
     def test_a_third_party_launcher_is_not_duplicated(self):
         user = "package:com.example.launcher\npackage:com.a\n"
         launcher = "com.example.launcher/.MainActivity"
@@ -239,17 +378,22 @@ class TestDiscoverySweep:
         by_pkg = {row["package"]: row for row in rows}
         assert by_pkg["ar.tvplayer.tv"]["name"] == "TiviMate"  # from the dump
         assert by_pkg["ar.tvplayer.tv"]["icon"] == "mdi:television-classic"
-        # a label-less app keeps its discovered name (the package id) and is
-        # still a row — it must not vanish from the table
-        assert by_pkg["com.example.handmade"]["name"] == "com.example.handmade"
-        # and every row's name/icon is a string
+        # t66: nothing can name `com.example.handmade` (the dump carries no label
+        # and it is not in the curated table), so its name is "" — the id is
+        # already published once in `package`, and a name that repeats it is the
+        # `pkg=pkg` row the user reported from the apps editor. The row itself is
+        # still offered: it must not vanish from the table.
+        assert by_pkg["com.example.handmade"]["name"] == ""
+        assert by_pkg["com.example.handmade"]["package"] == "com.example.handmade"
+        # every row has a string name ("" when nothing names it) and a real icon
         for row in rows:
-            assert isinstance(row["name"], str) and row["name"]
+            assert isinstance(row["name"], str)
             assert isinstance(row["icon"], str) and row["icon"]
 
     async def test_the_live_bug_the_discovered_list_is_a_superset(self):
-        """The complaint in one assertion: every app the user has, plus the
-        curated four, is offered — not just the curated four."""
+        """The complaint in one assertion: every app the user has is offered, and
+        each one is NAMED — on this device none of them is named by the firmware.
+        """
         from nsh.const import DEFAULT_APP_DIRECTORY
 
         poller = _DiscoveryPoller(dumps={"ar.tvplayer.tv": DUMP_TIVIMATE})
@@ -257,21 +401,19 @@ class TestDiscoverySweep:
 
         rows = await manager.async_refresh_installed_apps()
         packages = {row["package"] for row in rows}
+        by_pkg = {row["package"]: row for row in rows}
 
-        # the curated list the integration used to fall back to
+        # the curated table is exactly the app set the report listed
         assert set(DEFAULT_APP_DIRECTORY) <= packages
-        # ...and the apps it missed, which is what the user reported
-        for package in (
-            "ar.tvplayer.tv",                          # TiviMate
-            "com.github.damontecres.wholphin",         # Wholphin
-            "com.netflix.ninja",                       # Netflix
-            "com.google.android.youtube.tv",           # YouTube
-            "org.xbmc.kodi",
-            "com.plexapp.android",
-            "com.spotify.tv.android",
-            "com.example.handmade",
-        ):
-            assert package in packages, package
+        # ...and every one of those apps reads by its NAME, from the curated map,
+        # with the device naming nothing (t66)
+        for package, name in DEFAULT_APP_DIRECTORY.items():
+            assert name
+            assert by_pkg[package]["name"], package
+            if package != "ar.tvplayer.tv":  # this one's dump DOES carry a label
+                assert by_pkg[package]["name"] == name, package
+        assert by_pkg["ar.tvplayer.tv"]["name"] == "TiviMate"  # the device wins here
+        assert by_pkg["org.xbmc.kodi"]["name"] == ""  # known to nobody
         assert "com.google.android.tvlauncher" in packages  # + the launcher
         # a system app the device never lists as -3 is NOT invented here
         assert "com.android.systemui" not in packages
@@ -504,6 +646,13 @@ class TestDiscoveryUsesTheCachedLabelPath:
         dumps = [c for c in poller.shell_calls if "pm dump" in c]
         assert dumps  # the first sweep labelled everything it could
         assert len(dumps) >= 5  # ...and it did dump for the unlabelled ones
+        # t66: and then it STOPPED. On this device 5 probes in a row find no label,
+        # so the sweep does not spend one shell per package — the packaged-in
+        # in-flight window is the only slack above the probe limit.
+        from nsh.const import APP_LABEL_PROBE_LIMIT, APP_SWEEP_CONCURRENCY
+
+        assert len(dumps) <= APP_LABEL_PROBE_LIMIT + APP_SWEEP_CONCURRENCY + 1
+        assert len(dumps) < len(USER_PACKAGES.splitlines())
 
         # the cache now answers for EVERY package, so a forced re-sweep lists
         # again but dumps nothing at all (the 24 h label cache, not the table TTL)
@@ -514,11 +663,21 @@ class TestDiscoveryUsesTheCachedLabelPath:
         assert manager.app_label("ar.tvplayer.tv") == "TiviMate"
         by_pkg = {row["package"]: row for row in rows}
         assert by_pkg["ar.tvplayer.tv"]["name"] == "TiviMate"
-        assert by_pkg["org.xbmc.kodi"]["name"] == "org.xbmc.kodi"
+        # nothing names kodi (it is in no curated row, and the memo'd device says
+        # "no label"): the row shows the id ONCE, in `package`, and name ""
+        assert by_pkg["org.xbmc.kodi"]["name"] == ""
 
     async def test_a_failed_label_is_retried_by_the_next_sweep(self):
         """UNIT 0 semantics at scale: a transient failure is not permanent."""
-        poller = _DiscoveryPoller(dumps={"org.xbmc.kodi": "application-label: 'Kodi'"})
+        # a SMALL app set, with the failing package FIRST: it is the first probe the
+        # sweep makes, so its failure is recorded before the device-level memo can
+        # form — which is exactly the state the retry guarantee is about
+        packages = "package:org.xbmc.kodi\n" + "".join(
+            f"package:com.example.app{i}\n" for i in range(6)
+        )
+        poller = _DiscoveryPoller(
+            packages=packages, dumps={"org.xbmc.kodi": "application-label: 'Kodi'"}
+        )
         manager = _manager(poller)
         real_shell = poller.shell
         state = {"fail_kodi": True}
@@ -532,10 +691,19 @@ class TestDiscoveryUsesTheCachedLabelPath:
         poller.shell = _flaky
         rows = await manager.async_refresh_installed_apps()
         by_pkg = {row["package"]: row for row in rows}
-        assert by_pkg["org.xbmc.kodi"]["name"] == "org.xbmc.kodi"  # fell back
+        assert by_pkg["org.xbmc.kodi"]["name"] == ""  # fell back (nothing names it)
         assert manager.app_label("org.xbmc.kodi") is None  # NOT cached
 
+        # t66: the device-level memo is set by now (5 label-less answers), so the
+        # sweep will not fan out again — but kodi's FAILED probe is still owed its
+        # retry, and the app on screen is what drives it (t57 UNIT 0 survives).
+        assert manager.app_labels_absent is True
+        assert manager._probe_still_owed("org.xbmc.kodi") is True
+        assert await manager._dump_label("org.xbmc.kodi", on_demand=True) is None
+
         state["fail_kodi"] = False
+        await manager._fetch_app_label("org.xbmc.kodi")  # the on-demand retry
+        assert manager.app_label("org.xbmc.kodi") == "Kodi"
         manager._installed_apps = None
         rows2 = await manager.async_refresh_installed_apps()
         by_pkg2 = {row["package"]: row for row in rows2}
@@ -667,3 +835,174 @@ def test_importing_the_discovery_module_needs_no_home_assistant():
 
     assert module.__doc__ and "pm list packages -3" in module.__doc__
     assert not [n for n in dir(module) if n.startswith("hass")]
+
+
+class TestTheRealDeviceShape:
+    """t66: the user's own `pm dump com.netflix.ninja` shape, and what it implies.
+
+    Three properties are pinned against the fixture, because each one was a live
+    defect: the capture carries no label the parser can mistake for a name; the
+    fetch must not truncate the dump nor keep shelling every package once the
+    device has said it has no labels; and the name the user then SEES comes from
+    the curated table.
+    """
+
+    def test_no_label_is_extracted_from_the_whole_capture(self):
+        from nsh.ha_adb_common import parse_app_label, parse_application_label
+
+        # the capture's shape, in one place: a whole dumpsys (not a stub), whose
+        # package section is a resolver table rather than a manifest block
+        assert DUMP_NETFLIX_REAL.startswith("DUMP OF SERVICE package:\n")
+        assert "Activity Resolver Table:" in DUMP_NETFLIX_REAL
+        assert "DUMP OF SERVICE batterystats:" in DUMP_NETFLIX_REAL
+        assert len(DUMP_NETFLIX_REAL.splitlines()) > 60
+
+        # (a) NOTHING in it is a name — this is the "no label" the report saw, and
+        # it must be a parse result, not a truncation artefact
+        assert "application-label" not in DUMP_NETFLIX_REAL
+        assert parse_application_label(DUMP_NETFLIX_REAL) is None
+        assert parse_app_label(DUMP_NETFLIX_REAL) is None
+
+    def test_the_inferred_shapes_are_not_mistaken_for_names(self):
+        """The two ApplicationInfo lines the capture's record does not cover."""
+        from nsh.ha_adb_common import parse_label_res, parse_non_localized_label
+
+        assert parse_non_localized_label(DUMP_NETFLIX_REAL) is None  # =null
+        assert parse_label_res(DUMP_NETFLIX_REAL) is None  # a resource id
+        assert parse_non_localized_label("nonLocalizedLabel=Netflix") == "Netflix"
+        assert parse_label_res('labelRes="Netflix"') == "Netflix"
+        assert parse_label_res("labelRes=0x7f0f0d31") is None
+
+    async def test_the_search_runs_on_the_device_and_the_dump_is_never_truncated(self):
+        poller = _DiscoveryPoller(
+            dumps={"com.netflix.ninja": DUMP_NETFLIX_REAL}, on_device_grep=True
+        )
+        manager = _manager(poller)
+
+        # nothing names this package, and the probe still RAN: a definitive answer
+        assert await manager._dump_label("com.netflix.ninja") == ""
+
+        commands = [c for c in poller.shell_calls if "pm dump" in c]
+        assert len(commands) == 1
+        assert "head -300" not in commands[0]  # the defect: it cut the dump short
+        assert "grep -E" in commands[0]  # the fix: filtered before the wire
+        for shape in ("application-label", "nonLocalizedLabel", "labelRes"):
+            assert shape in commands[0], shape
+
+    async def test_one_shell_per_package_then_the_device_is_believed(self):
+        poller = _DiscoveryPoller(
+            dumps={"com.netflix.ninja": DUMP_NETFLIX_REAL}, on_device_grep=True
+        )
+        manager = _manager(poller)
+
+        rows = await manager.async_refresh_installed_apps()
+        dumps = [c for c in poller.shell_calls if "pm dump" in c]
+
+        # (b) the sweep did NOT spend one shell per package: on this device 5
+        # label-less answers in a row are the device-level conclusion, and only the
+        # packages already in flight can add to that count
+        from nsh.const import APP_LABEL_PROBE_LIMIT, APP_SWEEP_CONCURRENCY
+
+        assert len(dumps) <= APP_LABEL_PROBE_LIMIT + APP_SWEEP_CONCURRENCY + 1
+        assert len(dumps) < len(USER_PACKAGES.splitlines())
+        assert manager.app_labels_absent is True
+        assert "Netflix" in {row["name"] for row in rows}
+
+        # a forced re-sweep costs ZERO further probes: every package is cached
+        manager._installed_apps = None
+        await manager.async_refresh_installed_apps(force=True)
+        assert [c for c in poller.shell_calls if "pm dump" in c] == dumps
+
+    async def test_the_curated_name_is_what_surfaces(self):
+        """(c) the name the user sees, on every surface that shows one."""
+        from nsh.const import DEFAULT_APP_DIRECTORY
+        from nsh.ha_adb_common import resolve_package_name
+        from nsh.media_player import resolve_app_name
+
+        poller = _DiscoveryPoller(
+            dumps={"com.netflix.ninja": DUMP_NETFLIX_REAL}, on_device_grep=True
+        )
+        manager = _manager(poller)
+        rows = await manager.async_refresh_installed_apps()
+        by_pkg = {row["package"]: row for row in rows}
+
+        # the discovery table
+        assert by_pkg["com.netflix.ninja"]["name"] == "Netflix"
+
+        # the chain the entities call: no user map, and the device named nothing
+        assert resolve_app_name("com.netflix.ninja", {}, manager.app_label) == "Netflix"
+        assert (
+            resolve_package_name(
+                "com.netflix.ninja", device_label="", curated=DEFAULT_APP_DIRECTORY
+            )
+            == "Netflix"
+        )
+
+        # ...and the installed-apps sensor publishes the same table
+        sensor_module = load_entity_modules()["sensor"]
+        sensor = sensor_module.NvidiaShieldAppsSensor.__new__(
+            sensor_module.NvidiaShieldAppsSensor
+        )
+        entry = type("E", (), {})()
+        entry.entry_id = "e1"
+        entry.options = {}
+        entry.runtime_data = manager
+        entry.title = "mdarcy"
+        sensor.hass = entity_stublib.FakeHass()
+        sensor._entry = entry
+        assert any(
+            app["package"] == "com.netflix.ninja" and app["name"] == "Netflix"
+            for app in sensor.extra_state_attributes["installed_apps"]
+        )
+
+
+def test_the_name_chain_is_one_order_everywhere():
+    """user map > device label > curated > raw id, in ONE pure function (t66)."""
+    from nsh.ha_adb_common import resolve_package_name
+
+    curated = {"com.curated": "Curated", "com.both": "Curated Both"}
+    user = {"com.mapped": "Mapped", "com.both": "Mapped Both"}
+
+    assert resolve_package_name("com.mapped", user_map=user, curated=curated) == "Mapped"
+    assert resolve_package_name("com.curated", device_label="From Device") == "From Device"
+    assert resolve_package_name("com.curated", curated=curated) == "Curated"
+    assert resolve_package_name("com.both", user_map=user, curated=curated) == "Mapped Both"
+    # the device label sits BELOW the user map and ABOVE the curated table
+    assert (
+        resolve_package_name("com.both", user_map={}, device_label="D", curated=curated)
+        == "D"
+    )
+    # the id is shown ONCE, and an empty package has no name at all
+    assert resolve_package_name("com.nobody", curated=curated) == "com.nobody"
+    assert resolve_package_name("", curated=curated) is None
+    assert resolve_package_name(None) is None
+    # a blank device label is not a name (it is what a memo'd device returns)
+    assert resolve_package_name("com.curated", device_label="", curated=curated) == "Curated"
+    assert resolve_package_name("com.curated", device_label="  ", curated=curated) == "Curated"
+
+
+def test_the_device_label_lookup_is_lazy():
+    """A mapped app must not cost a device probe (t66).
+
+    ``media_player.resolve_app_name`` passes its ``app_label`` accessor, which
+    ENQUEUES an on-demand fetch on a miss: calling it for an app the user already
+    named would spend a shell to learn something nobody reads.
+    """
+    from nsh.ha_adb_common import resolve_package_name
+
+    calls: list[str] = []
+
+    def _device_label(package):
+        calls.append(package)
+        return None
+
+    assert (
+        resolve_package_name(
+            "com.mapped", user_map={"com.mapped": "Mapped"}, device_label=_device_label
+        )
+        == "Mapped"
+    )
+    assert calls == []  # the device was never asked
+
+    assert resolve_package_name("com.other", device_label=_device_label) == "com.other"
+    assert calls == ["com.other"]  # ...and it IS asked when the map misses
diff --git a/tests/test_media_sources.py b/tests/test_media_sources.py
index 329d438..adfbce0 100644
--- a/tests/test_media_sources.py
+++ b/tests/test_media_sources.py
@@ -57,7 +57,13 @@ class TestSourceList:
     def test_it_falls_back_to_the_curated_directory(self, modules):
         player, _served = _player(modules)  # no map configured
 
-        assert player.source_list == ["TiviMate", "Wholphin", "Netflix", "YouTube"]
+        # t66: the curated table grew to cover the apps the report listed, and the
+        # dropdown is the same table (so it is asserted against the table itself,
+        # not a copy that would drift)
+        from nsh.const import DEFAULT_APP_DIRECTORY
+
+        assert player.source_list == list(DEFAULT_APP_DIRECTORY.values())
+        assert "Netflix" in player.source_list  # a name the user recognises
 
     def test_a_blank_name_never_reaches_the_dropdown(self, modules):
         """t49: the map is hand-editable, so 'com.x=' must not offer an empty entry.
@@ -464,8 +470,25 @@ class TestCaseInsensitiveNames:
         assert player.source == "  TiviMate  "  # this surface shows it as typed
         assert player.source not in player.source_list
 
-    def test_a_blank_map_value_falls_through_to_the_raw_package(self, modules):
-        """Blanks are not displayed as nothing — the chain continues past them."""
+    def test_a_blank_map_value_falls_through_the_chain(self, modules):
+        """Blanks are not displayed as nothing — the chain continues past them (t66).
+
+        Past a blank map value comes the device label, then the curated table, then
+        the raw id: `ar.tvplayer.tv` is curated, so the dropdown row is its NAME.
+        The raw-id tail of the same chain is pinned in the test below.
+        """
         player, _served = _player(modules, app_map={TIVIMATE: ""}, foreground=TIVIMATE)
 
-        assert player.source == TIVIMATE
+        assert player.source == "TiviMate"
+
+    def test_a_package_nothing_names_stays_raw(self, modules):
+        """The last tier: no map value, no device label, no curated row → the id.
+
+        It is shown ONCE (the id itself), never as a `pkg=pkg` pair.
+        """
+        player, _served = _player(
+            modules, app_map={"com.example.unknown": ""}, foreground="com.example.unknown"
+        )
+
+        assert player.source == "com.example.unknown"
+        assert player.source_list == []  # a name is the ticker's raw id
diff --git a/tests/test_standalone_setup.py b/tests/test_standalone_setup.py
index 3e58bf4..1fc8389 100644
--- a/tests/test_standalone_setup.py
+++ b/tests/test_standalone_setup.py
@@ -1360,7 +1360,12 @@ class TestAppNameMapResolution:
         assert player.app_name == "TiviMate"
         assert player.app_id == "ar.tvplayer.tv"  # app_id stays raw
 
-    async def test_unmapped_package_stays_raw(self):
+    async def test_a_curated_package_reads_by_its_name(self):
+        """t66: an app the user did NOT map still gets a name — from the curated table.
+
+        This is the device the report describes: `pm dump` names nothing, and a
+        package absent from the user's own map used to read as its raw id.
+        """
         import entity_stublib
 
         modules = entity_stublib.load_entity_modules()
@@ -1371,7 +1376,21 @@ class TestAppNameMapResolution:
         manager._on_current_app("com.google.android.youtube.tv")
         await asyncio.sleep(0.05)
 
-        assert player.app_name == "com.google.android.youtube.tv"
+        assert player.app_name == "YouTube"
+
+    async def test_an_unmapped_uncurated_package_stays_raw(self):
+        """The chain's last tier, on the same surface: nothing names it → the id."""
+        import entity_stublib
+
+        modules = entity_stublib.load_entity_modules()
+        player, _s, _i = _cmd_player(modules)
+        player._entry.options["app_name_map"] = {"ar.tvplayer.tv": "TiviMate"}
+
+        manager = player._manager
+        manager._on_current_app("com.example.nobody")
+        await asyncio.sleep(0.05)
+
+        assert player.app_name == "com.example.nobody"
 
     async def test_cached_manifest_label_is_the_second_tier(self):
         """P5-1: below the explicit map sits the device manifest label."""