If a compactor is running when a shutdown is initiated, it attempts to shutdown quickly by not completing the compaction and must leave the file around just in case. Unfortunately, this can cause massive problems upon a restart as any attempt to compact this file again can cause all sorts of issues. For example if during compaction it is found to reduce the index count from 2 to 1 the next run the Compactor will reduce the count from 1 to 0 and remove that entry from the Index. In this case the value is still present in another data file but the index is incorrect!

It may be the only way to handle this properly is not to stop the Compactor from completing its current file and instead only avoid compacting new files. Then if this takes too long the user can force shutdown which we will in turn notice because the index is dirty and rebuild it, which means it will be fine.