In FX 2.x we use to get intermittent failures on Windows (and maybe Linux) when attempting to enqueue too many runnables. This is no longer a problem in FX 8 because we use the InvokeLaterDispatcher, but I verified that disabling it leads to problems, so I figured we could use some new unit tests.

I also discovered an issue with the way Util.runAndWait(Runnable...) waits, which causes each runnable to wait for up to the full timeout amount, which means that if you have, say, 10000 Runnables that hang, you will wait a very long time (50,000 seconds) before timing out. I need to fix this at the same time.