# Expected Diagnosis: tail-latency-surge-during-incident-window

This case should converge on a write-contention path rather than a generic
infrastructure explanation. The expected answer should identify a single-TiKV
hotspot driven by a repeated point-update workload, with direct lock-contention
evidence and commit/abort pressure during the incident window.

The expected evidence chain is:

- TiKV lock-related metrics show strong contention signals during the window.
- The hotspot is concentrated on one TiKV instance rather than evenly spread.
- Query evidence points to a repeated point-update write path as the dominant
  pressure path; the large join update is secondary.
- The final diagnosis links update frequency, lock contention, hotspot pressure,
  and tail-latency degradation.
- Recommended actions should target the write path, such as batching or
  asynchronous aggregation, instead of stopping at an infrastructure-only summary.