In the past, if I discovered that the pp file was computed for different
iterations, I removed everything from it.
Which is a bad idea, hence I'm not deleting the datasets that I know
have to be deleted.