These tests were always slow, but bots very clearly claim that they became ~50% slower at around r160908 (lazy DOM tree creation patch). I know this doesn't seem related, but the numbers appear conclusive.
Even if not a true regression, we should do something to make these tests not time out on bots.
It appears that these tests are over 30x slower in debug than in release - they don't show up on flakiness dashboard for release builds, which I believe means that they take under 1 second.
Skipped in http://trac.webkit.org/r162543 with Filip's blessing.
This doesn't happen any more.
These tests remain slow, although they shouldn't time out.
Another possible approach, given that all of the tests in question generate large arrays and preform computations on them, would be to simply decrease the size of the arrays the tests use.