Which jobs are flaking?

• pingcap/tiflow/pull_syncdiff_integration_test

Which test(s) are flaking?

• sync_diff_inspector/tests/sync_diff_inspector/checkpoint/run.sh
• The flaky assertion is in the bucket-checkpoint resume path, where the script picks the first resumed chunk by sorting lowerBounds only.

Jenkins logs or GitHub Actions link

https://do.pingcap.net/jenkins/blue/organizations/jenkins/pingcap%2Ftiflow%2Fpull_syncdiff_integration_test/detail/pull_syncdiff_integration_test/815/pipeline

Observed failure from build 815:
- resumed candidate logged as `39 upperBounds= indexCode=0:1-250:0:1`
- old assertion expected a fixed resumed index pattern and failed on `first_chunk_index`

Anything else we need to know

• Does this test exist for other branches as well?
  • Likely yes, if the same checkpoint test script is present.
• Has there been a high frequency of failure lately?
  • The same symptom has been seen in other PRs.
• Related parent tracking issue:
  • Subtask of Tracking issue for flaky tests #2246
• Root cause summary:
  • The checkpoint test assumes the first resumed chunk selected from logs is stable after sorting by lowerBounds only. In practice, multiple resumed chunks can share the same lower bound, so the script may pick a different valid resumed chunk and fail nondeterministically.