ack-timing

The response is sent before the work completes (or the work is not awaited), or slow work delays the response past the provider's timeout.

Why it goes wrong

Once the provider has a 2xx it will not retry, so failures after the acknowledgement lose the event; on serverless the function may be frozen mid-work. Conversely, a slow response makes the provider time out and retry while the first attempt is still running.

Safe pattern

Either do the work and then respond, or persist the event (or enqueue a job) and respond immediately, doing the work in a worker that has its own retries.

Configure

// retrysafe.config.ts
export default defineConfig({
  rules: { "ack-timing": "warning" }, // or "off"
});

// or, for one line (a reason is required):
// retrysafe-ignore ack-timing -- <why this is safe>