Replay and recovery

Bridge keeps the payload it rendered, so a replay hours later sends what happened, not what the database looks like now.

Three moves

Move	Endpoint	Use it when
Replay one delivery	POST /webhook/delivery/{reference}/replay	One event didn’t land, or your consumer crashed after acknowledging. Replaying a SUCCEEDED delivery is allowed.
Replay one event to every endpoint that received it	POST /webhook/event/{reference}/replay	The fault was on our side.
Read the payload back	GET /webhook/event/{reference}	You want the bytes without another delivery.

A replay creates a new delivery and never changes the original, so the record of what was attempted and what your server said can’t be edited. Replays are signed with your current secrets.

Reconciling a gap

1. GET /webhook/delivery?status=DEAD lists everything that ran out of attempts.
2. Fix the receiver, and confirm with POST /webhook/endpoint/{reference}/test.
3. Replay the dead deliveries, oldest first.
4. If your side has a gap that Bridge shows as SUCCEEDED, the loss happened after acknowledgement: GET /webhook/event/{reference} gives you the payload.