When a collection is dropped using the new drop-pending ident method in the KV storage engine, the collection is no longer visible to the rollback algorithm for estimating collection counts or writing rollback data files. The collection count will eventually be corrected during the replication recovery phase which is part of the rollback algorithm. For the rollback data file, it may be possible to provide access to the drop-pending ident through the storage engine interface but this is not in scope for the current work item in this ticket.