I haven't done intense tests yet, but based on my preliminary tests, the output is about 80% consistent. The others are like suggesting additional changes.