Investigate whether we can make vsync reliable in the presence of throttling

[cramforce]

E.g. we can see to always set a timeout for like 32ms and debounce that while rAFs fire.

This may be crucial for AMP in a box. CC @lannka, @dvoytenko

#7482 is an example of how this can go wrong.

[dvoytenko]

Certainly possible. We clearly separate in vsync animation frames vs measure/mutate frames. The non-animation frames can be enforced to execute with the timeout. Unfortunately, that would take some performance hit to run the "just in case" timer.

[lannka]

@cramforce I understand this causes problems if rAF is not reliable. But can you explain what's special about the inabox case?

[dvoytenko]

in-a-box iframes can and very likely will be throttled by browsers. Which mostly means that requestAnimationFrame will be either slow or won't arrive at all. Since we rely on rAF for measures/mutates, this is a significant issue.

[lannka]

It can be throttled for a "good reason" or by a bug (at least weird behavior). I think we only worry about the bug cases, right?

So we're worrying about that inabox is more likely to hit the bug (weird behavior)?

[dvoytenko]

It's not a bug - it's a browser feature. In-a-box is likely to hit it because it's likely to be out of viewport.

[cramforce]

I think the performance hit should be absolutely minimal