fix(ws): idempotent connection removal, release slot on any error

The status WebSocket pool removed connections with list.remove(), which
raises ValueError on a double-remove (broadcast already dropped a dead
socket, then disconnect tries again), and only released a slot on
WebSocketDisconnect — any other error leaked the socket into the
broadcast pool. Centralise removal in an idempotent _drop() called from
a finally block and from _broadcast.

ha-relevant: yes

backend/app/api/routes/status.py

@@ -1,3 +1,4 @@
+import contextlib
 import json
 
 from fastapi import APIRouter, WebSocket, WebSocketDisconnect
@@ -10,6 +11,12 @@ router = APIRouter()
 _connections: list[WebSocket] = []
 
 
+def _drop(websocket: WebSocket) -> None:
+    """Remove a connection if still present — idempotent, never raises."""
+    with contextlib.suppress(ValueError):
+        _connections.remove(websocket)
+
+
 @router.websocket("/ws/status")
 async def ws_status(websocket: WebSocket) -> None:
     # Accept first so we can send a close frame with a reason code
@@ -33,7 +40,11 @@ async def ws_status(websocket: WebSocket) -> None:
         while True:
             await websocket.receive_text()
     except WebSocketDisconnect:
-        _connections.remove(websocket)
+        pass
+    finally:
+        # Any error (disconnect or otherwise) must release the slot, else the
+        # dead socket lingers in the broadcast pool.
+        _drop(websocket)
 
 
 async def _broadcast(payload: str) -> None:
@@ -41,7 +52,7 @@ async def _broadcast(payload: str) -> None:
         try:
             await conn.send_text(payload)
         except Exception:
-            _connections.remove(conn)
+            _drop(conn)
 
 
 async def broadcast_status(node_id: str, status: str, checked_at: str, response_time_ms: int | None = None) -> None:

backend/tests/test_status.py

@@ -5,7 +5,13 @@ import pytest
 from fastapi.testclient import TestClient
 from starlette.websockets import WebSocketDisconnect
 
-from app.api.routes.status import _connections, broadcast_scan_update, broadcast_status
+from app.api.routes.status import (
+    _connections,
+    _drop,
+    broadcast_scan_update,
+    broadcast_service_status,
+    broadcast_status,
+)
 from app.main import app
 
 # ---------------------------------------------------------------------------
@@ -155,3 +161,63 @@ async def test_broadcast_no_connections():
     assert len(_connections) == 0
     await broadcast_status(node_id="n", status="online", checked_at="t")
     await broadcast_scan_update(run_id="r", devices_found=0)
+
+
+# ---------------------------------------------------------------------------
+# broadcast_service_status
+# ---------------------------------------------------------------------------
+
+@pytest.mark.asyncio
+async def test_broadcast_service_status_payload():
+    received: list[str] = []
+
+    class FakeWS:
+        async def send_text(self, text: str) -> None:
+            received.append(text)
+
+    fake = FakeWS()
+    _connections.append(fake)
+    try:
+        await broadcast_service_status(
+            node_id="node-7",
+            services=[{"port": 80, "protocol": "tcp", "status": "offline"}],
+            checked_at="2024-01-01T00:00:00",
+        )
+    finally:
+        _drop(fake)
+
+    msg = json.loads(received[0])
+    assert msg["type"] == "service_status"
+    assert msg["node_id"] == "node-7"
+    assert msg["services"] == [{"port": 80, "protocol": "tcp", "status": "offline"}]
+
+
+# ---------------------------------------------------------------------------
+# _drop — idempotent connection removal (regression for double-remove crash)
+# ---------------------------------------------------------------------------
+
+def test_drop_is_idempotent():
+    """Dropping a connection twice must not raise (was a ValueError crash)."""
+    class FakeWS:
+        pass
+
+    fake = FakeWS()
+    _connections.append(fake)
+    _drop(fake)
+    _drop(fake)  # second drop must be a no-op
+    assert fake not in _connections
+
+
+@pytest.mark.asyncio
+async def test_broadcast_dead_connection_dropped_once_safely():
+    """A send failure removes the dead socket without a double-remove crash."""
+    class DeadWS:
+        async def send_text(self, _: str) -> None:
+            raise RuntimeError("disconnected")
+
+    dead = DeadWS()
+    _connections.append(dead)
+    await broadcast_status(node_id="n", status="online", checked_at="t")
+    # A second broadcast must not raise even though dead is already gone.
+    await broadcast_status(node_id="n", status="online", checked_at="t")
+    assert dead not in _connections