fix(poller): F1 — the poller factory re-reads entry.options at call time (audit-sanctioned option), so the next built poller targets the NEW address after a zeroconf IP change; the stale live-read comment in config_flow points at the actual mechanism; TestIpChangeSelfHealF1 drives an IP change and pins the next poller host (t32)

custom_components/nvidia_shield/__init__.py

@@ -130,7 +130,18 @@ async def _build_poller_factory(hass: HomeAssistant, entry: ConfigEntry):
         )
         return None
 
-    return lambda: AdbMediaSessionPoller(adb.host, adb.port, adb.key_path)
+    # t32 F1 (review repair): the audit-sanctioned FACTORY RE-READ — the
+    # host/port resolve from entry.options AT CALL TIME, so the poller built
+    # after a zeroconf IP change targets the NEW address without a reload
+    # (the setup-time closure pinned the old address). A key-path change
+    # still needs the Pair ADB button / reconfigure reload, by design.
+    def _factory():
+        adb_now = resolve_adb_config(hass, entry)
+        if adb_now is None:
+            return None
+        return AdbMediaSessionPoller(adb_now.host, adb_now.port, adb_now.key_path)
+
+    return _factory
 
 
 async def async_setup_entry(hass: HomeAssistant, entry: ConfigEntry) -> bool:

custom_components/nvidia_shield/config_flow.py

@@ -363,7 +363,8 @@ class NvidiaShieldConfigFlow(config_entries.ConfigFlow, domain=DOMAIN):
                 if options != entry.options:
                     # t29 G3a: the deprecated update+schedule-reload combo is
                     # gone (2026.12 error) — the entry updates ONLY here; the
-                    # manager reads the host options live on the next cycle.
+                    # self-heal rides on the t32 F1 poller factory, which
+                    # re-reads entry.options at call time (no reload).
                     self.hass.config_entries.async_update_entry(
                         entry, options=options
                     )