Refactor BLE layer for 24/7 reliability

- Replace raw BleakClient with establish_connection from
  bleak-retry-connector (retries, GATT service cache, proxy-aware)
- Replace fragile asyncio.Event with asyncio.Queue for response frames,
  drain stale data on each connection to prevent cross-cycle leakage
- Register BLE advertisement callback to keep BLEDevice reference fresh
  across ESPHome proxy path changes
- Remove asyncio.sleep(2) device lookup hack
- Increase poll timeout floor from 10s to 20s
- Increase failure tolerance from 3 to 5 consecutive misses
- Bump default poll interval to 30s, min to 15s (halves connection churn)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

custom_components/xiaoxiang_bms/coordinator.py

@@ -5,6 +5,7 @@ import asyncio
 import logging
 from datetime import timedelta
 
+from bleak.backends.device import BLEDevice
 from homeassistant.components.bluetooth import async_ble_device_from_address
 from homeassistant.core import HomeAssistant
 from homeassistant.exceptions import HomeAssistantError
@@ -19,13 +20,13 @@ _LOGGER = logging.getLogger(__name__)
 # Only mark sensors unavailable after this many *consecutive* failed polls.
 # Transient BLE misses (device not in cache, ESPHome proxy busy, etc.) return
 # the last known data instead so the UI doesn't oscillate.
-_FAILURES_BEFORE_UNAVAILABLE = 3
+_FAILURES_BEFORE_UNAVAILABLE = 5
 
 
 class BmsCoordinator(DataUpdateCoordinator[dict]):
     """Polls the BMS over BLE and distributes data to all sensor entities.
 
-    Uses a connect → read → disconnect pattern on every poll so the BMS's
+    Uses a connect -> read -> disconnect pattern on every poll so the BMS's
     single BLE connection slot is free between updates (mobile app access).
     """
 
@@ -44,10 +45,12 @@ class BmsCoordinator(DataUpdateCoordinator[dict]):
         )
         self.address = address
         self._device_name = name
-        self._poll_timeout = max(poll_interval - 3, 10)
+        self._poll_timeout = max(poll_interval - 5, 20)
         self._handler = BmsBluetoothHandler(address)
         self.hw_version: str | None = None
         self._consecutive_failures = 0
+        # Kept fresh by the BLE advertisement callback registered in __init__.py
+        self._ble_device: BLEDevice | None = None
 
     # ------------------------------------------------------------------
     # Device info — shared by sensor, binary_sensor, number platforms
@@ -74,7 +77,7 @@ class BmsCoordinator(DataUpdateCoordinator[dict]):
 
     async def async_write_mos(self, value: int) -> None:
         """Send a MOS control command to the BMS, then refresh sensor state."""
-        device = async_ble_device_from_address(self.hass, self.address, connectable=True)
+        device = self._get_ble_device()
         if device is None:
             raise HomeAssistantError(
                 f"BMS ({self.address}) not reachable — cannot send MOS command"
@@ -82,9 +85,22 @@ class BmsCoordinator(DataUpdateCoordinator[dict]):
         success = await self._handler.write_mos(device, value)
         if not success:
             raise HomeAssistantError("BMS did not acknowledge the MOS command")
-        # Refresh immediately so sensors reflect the new MOS state
         await self.async_request_refresh()
 
+    # ------------------------------------------------------------------
+    # BLE device lookup
+    # ------------------------------------------------------------------
+
+    def _get_ble_device(self) -> BLEDevice | None:
+        """Return the freshest available BLEDevice reference.
+
+        Prefers the advertisement-callback reference (_ble_device) because it
+        tracks proxy transport path changes. Falls back to the scanner cache.
+        """
+        return self._ble_device or async_ble_device_from_address(
+            self.hass, self.address, connectable=True
+        )
+
     # ------------------------------------------------------------------
     # Poll
     # ------------------------------------------------------------------
@@ -92,7 +108,7 @@ class BmsCoordinator(DataUpdateCoordinator[dict]):
     def _handle_failure(self, reason: str) -> dict:
         """On a transient failure, return cached data up to the threshold.
 
-        Only raises UpdateFailed (→ sensors go unavailable) after
+        Only raises UpdateFailed (-> sensors go unavailable) after
         _FAILURES_BEFORE_UNAVAILABLE consecutive misses.
         """
         self._consecutive_failures += 1
@@ -109,14 +125,7 @@ class BmsCoordinator(DataUpdateCoordinator[dict]):
     async def _async_update_data(self) -> dict:
         """Connect to the BMS, fetch all data, disconnect."""
 
-        # The BMS may not be in the scanner cache immediately after a disconnect.
-        # Wait up to 2 s for an advertisement before giving up.
-        device = async_ble_device_from_address(self.hass, self.address, connectable=True)
-        if device is None:
-            await asyncio.sleep(2.0)
-            device = async_ble_device_from_address(
-                self.hass, self.address, connectable=True
-            )
+        device = self._get_ble_device()
         if device is None:
             return self._handle_failure(
                 f"BMS ({self.address}) not reachable — check Bluetooth adapter / proxy"