During rollback, replication writes out data files that contain documents that will be deleted after rollback completes. This means that the time it takes to execute a rollback scales linearly with the amount of data that has been inserted after the common point.

There may be cases when a user might not care about these deleted documents and simply want rollback to execute as fast as possible. We can add a startup-only setParameter that controls whether or not these rollback data files will be created or not; the default behavior will be to always write out data files.