[michaelmior]

> One way you might call modification time changes idempotent is if you don't care about modification times.

My point exactly :)