Aglet

How to learn from intermittent errors and recurring patterns

Intermittent-error learning turns scattered failures into a durable pattern record. Keep the request shapes, matched success, first divergent boundary, visible recovery, and uncertainty together. Then make the pattern easy to revisit when releases, data, timing, or dependencies change during future triage.

Keep the lesson for the next incident

  1. Record the failure taxonomy

    Document route, action, input shape, account or dataset class, statuses, error categories, retry outcomes, state risk, and release or timing conditions. Explain which dimensions were confirmed and which remained unknown. Avoid retaining sensitive values when structural fields reproduce the behavior.

  2. Preserve a matched outcome journey

    Keep a repeatable check for the original failure, a matched success, safe retry, and final state. Define which errors permit retry and which require correction or support. Assign ownership for interpreting a new occurrence instead of treating every similar status as the same problem.

  3. Recheck after shared-path changes

    Review the pattern after validation, transaction, dependency, retry, client state, or error rendering changes. Compare matched outcomes and state integrity again. Add a distinct case when a new route has different durability or recovery rules rather than broadening one intermittent test.

What to carry forward

The learning record should retain failure groups, matched successes, boundary evidence, retry ownership, state checks, and recurrence triggers. It is complete when future engineers can reproduce the pattern, state what remains uncertain, and distinguish a new reliability issue from a known one.

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