Liveness Check

Liveness probe — process + asyncio loop responsiveness check only.

Returns 200 + {"status": "alive", ...} if the FastAPI worker is reachable and the event loop is not wedged. Does NOT touch the database. Does NOT touch Redis.

Intended as the Traefik routing probe so that a sick database (pool drain, replica lag, query storm) does NOT cause Traefik to de-register the backend. When the DB is sick:

• /v1/health/live → 200 → Traefik keeps routing • /v1/health/ready → 503 → docker marks unhealthy, monitoring alerts fire, operator intervenes

The cached / static / Redis-only paths can still serve while the DB recovers, instead of the whole service 404’ing at the edge — which is exactly what bit prod in the 2026-05-17 outage (memory prod_outage_2026_05_17_entitlements_pool_drain).

The await asyncio.sleep(0) is intentional: it yields control back to the event loop for a single tick, so if a sync handler is pegging the loop the probe will still time out (correctly surfacing the wedge). Without it a wedged loop could in theory complete this coroutine synchronously between handlers.