Decide where the work belongs
Map work at risk
For each operation, record dependency, caller, local intent, current state, fallback age, and downstream decision. Separate display, report, write, and reconciliation paths. Attach real failure evidence and do not equate an HTTP error with outage scope.
Compare continuation choices
Assess pausing, stale read with label, local pending state, safe retry, readback, or owner review. Record freshness and side-effect limits. Do not return defaults as current data when the dependency normally establishes a required fact.
Set the outage queue
Queue unknown writes and decision-driving stale reads first, then shared dependency reach and blocked essential work. Set owner, review interval, recovery signal, and evidence threshold. Keep isolated optional reads behind a bounded scope.
What to carry forward
Return a dependency queue with failure class, work consequence, freshness, containment, reconciliation owner, and recovery proof. Escalate unknown writes and misleading fallback. Keep rank provisional when outage scope or recovery is unobserved. Keep stale-data risk visible in the queue.
Technical background: RFC Editor reference.
Keep the decision with the work.
Use a Work Item in Aglet to record the problem, the evidence you have, and the next decision. Add an owner and priority, then keep updates in the discussion so the next person can follow the reasoning.
Create an account See the product workflow