Intelligently reuse test results when isolating failures in batched PRs to save CI resources
When you use batching mode, the merge queue tests multiple PRs together for efficiency. If a batch fails, the queue needs to figure out which specific PR caused the problem by splitting the batch apart and retesting smaller groups.
With this optimization, the merge queue will recognize when it's about to run a test that it has already run during the isolation process. If we already know a test will fail based on earlier results, we skip rerunning it and reuse that information.
This means faster failure isolation when batches break, significant CI cost savings during the bisection process (especially important for large batches or expensive test suites), and quicker feedback about which PR needs fixing.