I didn't quite understood your proposal. So basically you are saying whenever we hide the mouse cursor we should treat it as if it moved outside of the page completely and when it appeared somewhere it is as if it came from outside of the page and we should send boundary events in either case accordingly. Is that correct?
Wasn't there a bug (I couldn't find it) that the login flow of one of the websites were broken due to sending mouse leave when mouse hides or am I mistaken?

Ah yes you are correct; we don't send the mouse leave for interop purposes. But what is weird is we update the hover state of pages that lay out and it shows what the last hovered node is.

So yeah. I'm totally up for matching current hover state update and also send the boundary events in favor of consistency and predictability. I'm not sure what the breakage will be in that case though.

Marginally related, that we also remove hover state on mac when mouse opens the context menu which I'm again in favor. But I recall some people were complaining about.

Closing as this is a can of interop worms and there is no immediate need.