Service workers & promises—keeping async work alive

TL;DR

• Use event.waitUntil(promise) for every async path started by an event.
• Design idempotent jobs—assume retries.
• Keep heavy/long jobs off transient events; use queues/alarms.

Core patterns

• Warm start (lazy imports), bounded retries with backoff, consistent error shapes.
• Message orchestration: command envelopes with requestId.

Don’ts

• Long-lived state in globals; the service worker can terminate at any time.
• Fire-and-forget promises from listeners.

Testing

• Simulate termination and replay; inject artificial delays; verify retry limits.

See also: Messaging Patterns · Talk to an Engineer