#721 Workers become zombies after asyncio crash - container healthy but all threads dead

OBSERVED: Workers can crash with asyncio.exceptions.CancelledError during HTTP operations, causing all internal threads (TimeoutService, task processing loop, etc.) to die while the container remains 'running' and 'healthy'. Symptoms: - Container status: running/healthy - Zero log activity - Stuck workflows not cleaned up - TimeoutService not running Root cause: asyncio task group error during HTTP connection: asyncio.exceptions.CancelledError: Cancelled via cancel scope ... Impact: Workflows get stuck in 'running' state indefinitely because TimeoutService (which cleans up expired claims) is dead. Workaround: Restart workers with 'docker compose restart worker' Suggested fix: Add a watchdog that monitors thread activity and restarts worker if all threads are dead. Or improve asyncio error handling to prevent cascade failures.