I found it was cleaner to run it with -a (autocorrect) enabling one cop at a time between commits. Less giant patch storm of everything, and more coherent changesets.