Earlier quoted context omitted.
A general golden UX rule is to never cause layout shifts on hover events, preventing this sort of thrashing. Unfortunately, Google/YouTube and others have lost the plot and no longer employ good UX professionals in their flagship products.
I really wish UIs would stop doing things on hover events altogether. Just because I stop my mouse doesn't mean I intend to do anything, let alone auto-play a video or have some pop-up explain something to me. Let me rest my goddamn hand for a second!
But yeah, hover is just for info about the thing you're about to click, not pretending I've already clicked it when I haven't.
The stupid thing about Youtube is that sometimes a video starts playing on hover, in the thumbnail and without sound, and then I actually do click it because indeed it is something I want to watch, and then I miss the first couple of seconds because it's already shown those, with the sound off. It's just the most bizarrely stupid UI decision I can think of.