We can run into this scenario:

Within the stageInsertBatch/stageInsertBatchIntoEligibleBucket path, we mark a rollover action as kHardClose due to schema change.

We then rollover it in findAndRolloverOpenBuckets without calling updateRolloverStats.
This will then lead to incorrect numBucketsClosedDueToSchemaChange. Or any hard-closed metric.
I believe this is the only place where we directly rollover without updateRolloverStats; we may also need to take a look at the abort behavior and see if this is consistent with the old write path/if we would need to updateRolloverStats at any point.