# Expected Diagnosis: insert-latency-degradation-during-incident-window

The expected result should converge on the case-specific mechanism below while
keeping conclusions bounded by the collected evidence.

## Expected conclusion

DNS resource exhaustion first delayed name resolution and connection establishment. Application-side query kills then exercised an Auto ID service cancellation bug, leaving Auto-ID-related inserts abnormally slow after the DNS event had eased, especially on a newly started TiDB instance. DNS capacity and the Auto ID hotfix are both required; treating the whole window as a single storage slowdown is incorrect.

## Required evidence chain

- Respect the corrected incident timeline and show that DNS pressure begins the event but does not explain the full duration.
- Correlate application query kills/context cancellation with continued Auto-ID-related INSERT latency.
- Identify the Auto ID cancellation bug and the concentration on the newly started TiDB instance as the persistence mechanism.
- Recommend DNS resource expansion plus the Auto ID hotfix or upgrade, and review application kill/retry behavior.