Apparently mrvdb has problems with duplicate inserts and missing files when unlinking. It could be due to coding, or it could be due to parallelizing.