Recovery Timeouts
Understand Beam transfer recovery windows and participant obligations.
BeamCore treats each task offer as participant-owned once its assignment wave has been dispatched. Each assignment wave has a 15 second timeout that starts after the full dispatch wave settles.
When the timeout fires, unfinished work is reassigned through the same participant-neutral PRISM allocation path. Every other recovery types use that same reassignment path immediately.
Orchestrator obligations
- Keep the NATS control connection and worker gateway healthy.
- Handle the full delivered task-offer wave.
- Forward offers and results immediately.
- Retain each result until BeamCore returns a terminal
task_result_ack.
Worker obligations
- Queue every valid offer and execute it as capacity becomes available.
- Report success and failure through
task_result. - Keep active tasks moving and report results before the assignment timeout expires.
Stalled or failed work is reassigned and can reduce future PRISM routing share.