If we fail InitialSync, we clean up/ reset the catalog before retrying InitialSync. If this occurs, it is safe to allow for untimestamped writes throughout _truncateOplogAndDropReplicatedDatabases. We already allow for untimestamped writes: here. The issue is that this flag gets reset after one WUOW/transaction commits or aborts.
Proposed solution: add an RAII type to allow for untimestamped writes over a chunk of code - resilient to being reset upon wuow commit/abort on the same RecoveryUnit.