Watch an async pipeline survive chaos — live.

A real job queue streamed over SSE: 3 workers, exponential backoff with jitter, max-3 retries, dead-letter. Run the chaos scenario — the provider starts failing 65% of calls, a worker dies mid-job — and watch the system absorb it instead of losing data. This is the state machine behind every reliable Celery / BullMQ / RQ deployment.