One issue here is there's no exact test case. As a result, all variants seem to use slightly different timing curves. Maybe that's irrelevant, but it's at least annoying to look at.

(Not to say I don't appreciate this page.)