Decide where the work belongs
Classify the consequence
For each response, identify operation, status, durable effect, user-visible recovery, and whether a retry already occurred. Distinguish unknown body from known permanent error. Attach real examples and avoid treating status code frequency as impact without the operation context.
Compare safe containment
Consider stopping retries, preserving the raw response safely, using an existing typed fallback, or asking for contract clarification. Record what each choice leaves pending. A generic 500 branch may protect a write while making read recovery less useful, so state the tradeoff.
Choose the error queue
Set owner, review time, decoder or contract dependency, and evidence threshold. Queue unknown write outcomes and shared parser breaks first. Keep copy or logging improvements behind an error mapping that makes the underlying action safe.
What to carry forward
Return an error queue with response class, operation consequence, retry exposure, safe containment, owner, and proof needed to reprioritize. Escalate unknown writes and shared decoder paths. Mark the ranking provisional where the remote error contract or intermediary response is unverified.
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