Fix health check loop silently dying on uncaught exception

Wrap check_all() call inside the loop with try/except so a transient error
cannot exit the while-True and freeze all health statuses. Add transition
logging (HEALTHY / UNHEALTHY) so docker logs show when a service changes
state. Also add refetchIntervalInBackground on the frontend query so the
poll continues even when the browser tab is not focused.

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

backend/app/services/service_health.py

@@ -7,7 +7,7 @@ no blocking calls on the request path.
 """
 import asyncio
 import logging
-from dataclasses import dataclass, field
+from dataclasses import dataclass
 
 import httpx
 
@@ -62,6 +62,7 @@ def register_services(doc_service_url: str, ai_service_url: str) -> None:
     ]
 
     _health = {svc.id: None for svc in _REGISTRY}
+    logger.info("Service registry initialised with %d services", len(_REGISTRY))
 
 
 # ── Health check logic ────────────────────────────────────────────────────────
@@ -69,12 +70,23 @@ def register_services(doc_service_url: str, ai_service_url: str) -> None:
 
 async def _check_service(svc: ServiceDefinition) -> None:
     url = f"{svc.internal_url}{svc.health_path}"
+    prev = _health.get(svc.id)
     try:
         async with httpx.AsyncClient(timeout=5.0) as client:
             resp = await client.get(url)
-        _health[svc.id] = resp.status_code == 200
+        healthy = resp.status_code == 200
-    except Exception:
+    except Exception as exc:
-        _health[svc.id] = False
+        logger.debug("Health check failed for %s: %s", svc.id, exc)
+        healthy = False
+
+    _health[svc.id] = healthy
+
+    # Log only on transitions so the logs stay quiet during normal operation
+    if prev != healthy:
+        if healthy:
+            logger.info("Service %s is now HEALTHY", svc.id)
+        else:
+            logger.warning("Service %s is now UNHEALTHY", svc.id)
 
 
 async def check_all() -> None:
@@ -83,9 +95,16 @@ async def check_all() -> None:
 
 
 async def health_check_loop() -> None:
-    """Runs forever; polls every POLL_INTERVAL seconds."""
+    """Runs forever; polls every POLL_INTERVAL seconds.
+
+    Exceptions inside a single polling round are caught so the loop cannot
+    be killed by a transient error.
+    """
     while True:
+        try:
             await check_all()
+        except Exception:
+            logger.exception("Unexpected error during health check round; will retry")
         await asyncio.sleep(POLL_INTERVAL)
 
 

frontend/src/pages/AppsPage.tsx

@@ -32,6 +32,7 @@ export default function AppsPage() {
     queryKey: ["services"],
     queryFn: getServices,
     refetchInterval: 30_000,
+    refetchIntervalInBackground: true,
   });
 
   return (