Prove booking recovery after a lost confirmation response
Checkout commits successfully, then the connection closes. A customer retries while the expiry worker is also running.
- Focused work estimate
- 5h + prerequisites
- Priority in the scenario
- High
- Engineering practice
- Fault injection · Recovery
Estimated field mix
- Quality engineering40%
- Backend30%
- Distributed systems30%
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
A fictional equipment service books maintenance visits. Expiring checkout holds and reschedules occasionally double-book a technician.
Setup prerequisites
- Build synthetic technicians, slots and booking records.
- Use an injectable clock and separate clients for concurrency probes.
Preceding work
Complete these dependencies, or supply their agreed outputs before taking this ticket.
- ABOOK-101 · Normalize maintenance-slot requests to explicit UTC intervals
- ABOOK-102 · Model overlapping technician availability as half-open intervals
- ABOOK-104 · Reserve one active hold for a contested maintenance slot
- ABOOK-105 · Expire holds without cancelling a booking confirmed at the boundary
- ABOOK-106 · Reschedule a visit without releasing its old slot prematurely
- ABOOK-108 · Detect orphan holds without changing confirmed appointments
- ABOOK-107 · Calculate the cancellation window from the booked service policy
- ABOOK-109 · Hide customer details in scheduling conflict responses
Acceptance criteria
- Document persisted state at each interruption point.
- A stable command key resolves the committed confirmation.
- No retry creates a second booking or frees its slot.
Implementation constraints
- Build a deterministic fault-injection harness using synthetic visits.
Verification to include
- Drop the response after commit and retry to the same booking.
- Drop before commit, run expiry, and verify the declared expired outcome.
Deliverables
- Recovery trace and executable interruption probe
Rollout and recovery
Run the drill before enabling checkout; fall back to read-only booking lookup during incidents.
Value of the work
For the engineer: Practice resource allocation, interval conflicts and durable workflows.
For the team: Review scheduling correctness and recovery before exposing scarce capacity.
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.