In REP-3572 we saw a mongosync test failure due to awkward interaction between mongosync's "fruit collections" substitution logic and time-series collection TTLs.

Along the way, I saw something weird: the destination cluster in that test run has these oplog events for a time-series buckets collection:

• create
• rename
• delete

Note that there's no insert here.

I've inquired to a few folks, all of whom seemed to regard this as strange behavior. It seems worth confirming that nothing actually wrong happened here.

See REP-3572 for more details.