OK. well that sounds reasonably positive I guess - FWIW the 200ms delay I set was entirely arbitrary, for the purposes of testing. It may be (should be) feasible to shorten that and still get robust behaviour, but lacking a suitable X11 box to try it on today, I just picked something...

 

OK, that's a pity: What that's really telling us is that I implemented it wrong, as that approach (or at least something very like it, only better!) should work!