Retries and recovery
The retry schedule, disabled endpoints, recover and replay, and the deliveries log.
curl "$SEQ/v1/webhooks/wep_01J9…/deliveries?limit=50" -H "Authorization: Bearer $ADMIN_KEY"
curl -X POST $SEQ/v1/webhooks/wep_01J9…/recover -H "Authorization: Bearer $ADMIN_KEY" -H "Content-Type: application/json" -d '{"since":"2026-10-12T00:00:00Z"}'
curl -X POST $SEQ/v1/webhooks/wep_01J9…/enable -H "Authorization: Bearer $ADMIN_KEY"
curl -X POST $SEQ/v1/webhooks/deliveries/wdl_01J9…/replay -H "Authorization: Bearer $ADMIN_KEY"The schedule
A delivery is attempted at once, then after each failure with this backoff, about three days in total:
| Attempt | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 |
|---|---|---|---|---|---|---|---|---|---|
| Delay before it | 0 | 5 s | 5 min | 30 min | 2 h | 5 h | 10 h | 10 h | 10 h |
Each attempt waits up to 10 seconds for your response. A 2xx ends it (delivered); after the ninth failure the delivery is failed for good and only a recover or a replay sends it again. The delivery row records every attempt, the last status code and the last error.
Disabled endpoints
An endpoint whose deliveries have all failed for three days (the oldest pending delivery is older than that) is set to disabled, with disabled_reason and disabled_at; its remaining deliveries are held, not lost; and endpoint.disabled is delivered to the project's other endpoints. An endpoint whose signing secret can no longer be decrypted (the encryption key was removed from SEQ_ENCRYPTION_KEYS) is disabled the same way: rotate the secret, then enable.
POST /v1/webhooks/{id}/enable sets it back to active and re-queues every held delivery, oldest first. Nothing was lost.
Recover and replay
POST /v1/webhooks/{id}/recover { since }re-sends every delivery of the endpoint that failed for good since that time, with a fresh attempt counter.POST /v1/webhooks/deliveries/{id}/replayre-sends one delivery whatever its state (delivered, failed, pending), with a fresh counter. The endpoint must be active (409 ENDPOINT_DISABLED).
A replayed delivery keeps its webhook-id, so an idempotent receiver ignores a duplicate.
The deliveries log
GET /v1/webhooks/{id}/deliveries (paginated, updated_since) returns, per delivery: id, event_id, event_type, enrollment, status (pending, delivered, failed, disabled), attempts, next_attempt_at, last_status_code, last_error, payload, created_at, delivered_at. Delivered rows are pruned after 30 days.
Ordering during failures
Deliveries for the same endpoint and enrollment go one at a time in creation order: while one is retrying, the ones behind it wait (waiting), and go out as soon as it succeeds or gives up. Other enrollments and other endpoints are not affected.
What to alert on
An endpoint with status: "disabled", a delivery with attempts >= 5, or last_error repeating across deliveries. The operator's GET /v1/admin/engine/health and the project's GET /v1/engine/health show the seq_webhooks queue depth.