Geldstuck retries webhook deliveries until your endpoint returns 2xx or we exhaust the schedule. You’ll never silently lose an event - failures are surfaced in the dashboard and on a webhook.delivery.failed meta-event.

​
Retry schedule

AttemptDelay before attempt
1Immediate
22 seconds
34 seconds
416 seconds
564 seconds
64 minutes
716 minutes
864 minutes
94 hours
1016 hours
Total retry window: ~24 hours. After the 10th failed attempt, the event is moved to the dead-letter log.

​
What counts as failure?

  • Any non-2xx response.
  • Connection refused / DNS failure.
  • Handler doesn’t respond within 10 seconds.
  • TLS errors (expired cert, hostname mismatch).

​
What counts as success?

  • Any 2xx response, with any body.
  • Including 204 No Content.

​
The dead-letter log

Every tenant has a dead-letter view in the dashboard under Developers → Webhooks → Dead letter. Events here:
  • Stopped retrying after 10 attempts.
  • Stay queryable for 30 days.
  • Can be replayed manually (see below).

​
Replaying events

From the dashboard, select any event in the delivery log - live or dead-lettered - and click Replay. The same event is re-delivered; your handler should handle it idempotently. Programmatically: 
POST /v1/webhook-endpoints/:endpointId/replay
{ "eventId": "evt_01HX3ZJ..." }
Or replay a range (useful after an outage): 
POST /v1/webhook-endpoints/:endpointId/replay
{
  "from": "2026-04-22T09:00:00Z",
  "to":   "2026-04-22T11:00:00Z",
  "types": ["kyc.*"]
}
Replayed events are marked with X-Geldstuck-Replayed: true. If your handler needs to behave differently on replays (e.g., skip outbound email), branch on this header.

​
Idempotency

Because we retry, your endpoint will occasionally see duplicates. Defend with the event ID: 
const handled = await redis.set(
  `gs:event:${event.id}`,
  "1",
  { NX: true, EX: 60 * 60 * 24 * 7 },  // 7-day TTL
);
if (handled !== "OK") return;  // already processed
// ... do the work

​
Monitoring

  • The Deliveries view shows live success/failure counts, p50/p95 latency, and status-code distribution.
  • Subscribe to webhook.delivery.failed (a meta-event) to get alerted in your existing ops pipeline - it fires on the 3rd consecutive failure of an endpoint.
  • Endpoints that fail > 95% of deliveries for 24 hours are auto-paused and flagged in the dashboard; we’ll email the account owner before it happens.