A status URL PagerDuty can poll directly

Your worker pool fell over twice last quarter and nobody noticed for forty minutes. You needed a health check. The old options: write a /healthz route in every service, deploy a sidecar, install Datadog agents, configure Prometheus.

GET /api/v1/daemon/status/{worker_id} returns JSON with status: "RUNNING" and a PID. PagerDuty's HTTP integration polls it every minute. The moment status flips to BACKOFF or FATAL, the alert fires.

You did the same for Postgres, Redis, the queue consumer, the cron service. Six daemons, six status URLs, one polling rule per URL in PagerDuty. No agent installed in the container. No exporters. No scrape config. The supervisord state machine is the source of truth, exposed as HTTP.

The health check stopped being a script you run. It became a URL you poll.