Reconcile a task claim when its acknowledgement is lost
A responder claims Investigate cache saturation, loses the acknowledgement, and retries. The room appends duplicate claims and briefly shows two owners.
- Focused work estimate
- 2h 30m + prerequisites
- Priority in the scenario
- High
- Engineering practice
- Idempotency · Transactions · Optimistic concurrency
Estimated field mix
- Real-time systems40%
- Backend30%
- Database engineering30%
Field percentages are editorial estimates of the ticket's engineering focus. They total 100%; they are not measured time, proficiency scores, or ownership evidence.
Review it, then add it to your workspace.
The board opens an editable draft; nothing is saved until you confirm it. Sign-in and workspace permissions apply, and Demo boards remain ephemeral.
Project context
The fictional Harbor operations team coordinates synthetic incidents in a shared room. Engineers post status notes and claim response tasks while connections come and go. Scope is one application gateway, a durable room event store, and fixture clients; alert paging and external chat delivery are excluded.
Setup prerequisites
- Room membership and command contracts
- Synthetic incident events and a controllable transport harness
Preceding work
Complete these dependencies, or supply their agreed outputs before taking this ticket.
- LIVE-101 · Show the difference between connected and caught up
- LIVE-102 · Reject malformed room events before they reach the reducer
- LIVE-103 · Authorize room subscription before replaying its history
- LIVE-104 · Stop duplicate and out-of-order notes confusing the timeline
- LIVE-105 · Recover when a reconnect cursor is older than retained history
Acceptance criteria
- A claim command has a stable operation key and expected task revision.
- The durable event and operation result commit together or neither becomes authoritative.
- Retry returns the existing result; a competing confirmed claim produces a conflict instead of a second owner.
Implementation constraints
- Use the application transaction boundary and an outbox if broadcasting crosses processes.
Verification to include
- Accept a claim, drop the acknowledgement, and retry with the same key; one logical claim exists.
- Race two responders at the same revision and fail broadcast after commit; one owner remains and replay recovers the event.
Deliverables
- Idempotent claim command and lost-acknowledgement/concurrent-claim tests
Rollout and recovery
Enable claims after replay recovery; disable new claims if transaction/result consistency fails while keeping confirmed assignments readable.
Value of the work
For the engineer: Practice event identity, ordering, replay, authorization, and bounded backpressure in one understandable system.
For the team: Inspect how an engineer keeps collaborative state trustworthy during failures without relying on optimistic success messages.
Evidence boundaries
Outcome Evidence: Tests, patches, and runbooks are requested deliverables. They become Outcome Evidence only through a qualified Mission and immutable Evidence IDs.
Ownership Evidence: Independent adaptation must be observed under a declared verification policy and cite immutable Evidence IDs. Completing a planning ticket establishes no Ownership Evidence.