There are a couple of problems with the current implementation to report the "oplogEnd" timestamp.

1. The timestamp is fetched outside the critical section where we ensure no checkpoint took place. This can result in the "oplogEnd" timestamp being reported as less than the checkpoint timestamp the backup cursor was opened against.
2. The "oplogEnd" timestamp is not majority committed. It is the last applied optime, and can be rolled back.

If we remove this, it would only be in the $backupCursor and not $backupCursorExtend.