Currently updateCallbackDone is a reaction to the provided update callback (in startViewTransition(update)),
and activating the view transition is a reaction to updateCallbackDone.

This makes it so that author-provided updateCallbackDone reactions are usually called before capturing the new state, but sometimes after. This can create a confusion and subtle bugs.

I found that updateCallbackDone is most useful in its usual behavior, when all the author reactions are flushed and only then the new state is captured - this allows using this callback as a way to add new state changes on top of whatever is in the update callback

Proposing to slightly change the wording so that it's clear that there's a microtask checkpoint between resolving updateCallbackDone and activating the view transition (capturing the new state).