Just because it isn't public doesn't mean no one calls it :).  But seriously, that's why I said "the risk".  Is the small benefit of the codechange worth the small risk of breaking something?  I'm not answering that question, I'm posing it.  Call me -0 on the change.

I agree that the optimization argument doesn't really seem relevant.  If someone cares about performance they aren't going to be using pathlib.