So I thought tresize was going to be the easiest thing to profile, since we should be able to see the difference consistently.
One of the things that stick out is that we still spend time updating scrollbars even in the
overflow: hidden case, because we always create scrollbars even for
overflow: hidden on the root... Not 100% sure that will account for all the regression though, scrollbars seem like a fairly small part of the page... Bug 1590247 would probably make this unnecessary, as the only reason we always create scrollbars on the root is to handle
overflow: auto ->
overflow: hidden more nicely, and bug 1590247 would make the optimization much more generic.
cc'ing some folks which may want to take a look at the profile and may have more insights.