Receive and process a provider notification
What It Does
Ingress retains provider material before acknowledgement, then queues signature verification and interpretation. Accepted messages produce typed events for applications; duplicate or rejected messages remain inspectable. Replay creates separate message records from retained bodies.
The Flow
When Something Fails
The provider controls retries after request failure. No Rhythm record is guaranteed before ingress succeeds. Reject missing endpoints or oversized input; permit-store failure admits the request. Provider retries can generate another ingress request. Return a storage or record error before receipt acknowledgement. Successfully stored objects can remain orphaned until expiry. Retry publication from the retained dispatch row. A crash between publication and deletion can duplicate queue delivery. Retry classified transient failures up to six attempts and stop permanent failures immediately. The received message remains inspectable with its failure outcome. Rollback failed event recording and leave processing to recovery. No subset of the interpreted events commits alone. A replay whose stored body expired ends dead. Replayed events can reach consumers that handled earlier processing.
Open Design Questions
Product still needs an accepted message milestone and an explicit replay-consumer contract. Stored-body expiry limits recovery.