Failure Mode Classification

A report such as "sometimes X is stale" gets re-investigated from scratch when it returns. A stable label — "failure mode 2: retries exhausted with no terminal write" — links occurrences and makes the fix PR a starting point for the next investigation.

Name the failure mode before fixing it

When a bug is one of several plausible ways the same system can fail, give it a stable name and number, tracked against a parent issue or taxonomy — not just a description in the fix PR. "Failure mode 2" only works as a label if there's a "failure mode 1" it's distinguished from, and a place (the parent issue) where both are listed together. This turns "have we seen this before?" into a lookup instead of an investigation.

Root-cause to mechanism, not symptom