I started looking at this in SERVER-79610 but it will need a deeper look.

There are some things I found surprising with timeseries even without bulkWrite:

• Timeseries with 1 retry seems to increment the arrayFilter metrics twice but that might be a test issue
• document.updated is not increased on a timeseries update, neither is metrics.queryExecutor.scanned (on update or delete).
• timeseries with retry do not increase writeConcern.update.majority, but writeConcern.update.none
• timeseries insert does not increase writeConcern.insert.majority

On top of those bulkWrite + timeseries has its own issues:

• Timeseries with retry and bulkWrite does not increment the arrayFilter metrics twice but that might be a test issue (unlike timeseries with retry and no bulkWrite)
• document.updated with timeseries behaves differently between no bulkWrite (and bulkWrite with retry) and bulkWrite with no retry.
• Timeseries with bulkWrite and retries does not increase retriedCommandsCount or retriedStatementsCount, a quick look seems it does not see a retried stmtId at all so need more investigation.

For now I stashed the changes to the test that allowed me to uncover those. I will go back to this after I finish the SERVER-79610 PRs.