How long does Sume retry a webhook before giving up?

Not long: your receiver can be down for about 4.5 minutes of retries on a job webhook, or about three hours on a run webhook, before Sume stops trying automatically. After that, one call to redeliver or one poll recovers the event.

Timings are my arithmetic from Job webhooks and Run webhooks, read 2026-09-30.

How does that compare with the Standard Webhooks advice?

The spec recommends a schedule spanning multiple days, with exponential backoff and jitter. Neither Sume schedule reaches that, so size your deploys and outages against the numbers below rather than against a multi-day assumption.

How much downtime can each webhook ride out?

Ten attempts leave nine gaps. Job webhooks use a 30-second default gap. Run webhooks use the docs' formula min(max(30s x 2^(attempt-1) with jitter, Retry-After), 1h).

What counts as down?

Anything that is not a 2xx inside 10 seconds: timeouts, errors, and redirects all consume an attempt. A restart that takes longer than the window loses all ten attempts, even if the app is healthy again a minute later.

How do I recover a missed event?

Call POST /v1/jobs/{job_id}/webhook/redeliver (needs jobs:write). It re-sends the real terminal event with a fresh signature and works after the automatic attempts are exhausted. Or poll status_url, since the job reached its terminal state regardless. A walkthrough of the per-attempt rules is in the ten-attempt schedule.