It seems to me that supporting multiple fullscreen element in a document is one of the main sources of complexity of Fullscreen API, and I'm wondering whether this complexity is really necessary. I don't really see how useful it is to allow fullscreen multiple elements, and I suspect whether it is used in wild.

Probably we can try putting a use counter and see whether there is really any use of nested fullscreen, and if not, we should probably remove it. Thought?