Code review comment for lp:~kdub/mir/display-groups

(3) Even if you have multiple displays running at the same rate, they will be out of phase by half a frame on average, sometimes worse. So if you synchronize their post() calls then you're dooming all-but-one display to get a shorter render time than 16ms, and this will often leave insufficient render time to composite a whole frame without missing the next one. Our old workaround of triple buffering saves us here, but that's a step backwards in performance. So you have to choose between high lag or sometimes skipping frames. For this reason, I disapprove of the groups design.
It's important each display's flip() occurs on a separate thread asynchronously, so that its own rendering of the next frame can start immediately and we have a full frame time to render.

We need to reduce our use of grouping logic (bug 1395416) rather than extending it.

If you still need some sort of grouping to resolve Android problems, I suggest utilizing the grouping logic ("overlapping outputs") we already have, and only utilizing it in Android.