I have wondered if they're combining them on the server-side and just pushing a single stream to the client. Would that be detectable?
They may have some good video-compression stuff going on. I haven't bothered to set up a network sniffer to see what is going on, but I suspect that it's a PtP vid connection.
This is because I will occasionally see only one thumbnail bork, and also the thumbnails are constantly switching places to match my screen (which I find maddening).
I think that it was Skype that had the "lowest common denominator" problem, where just one bad connection would bork the whole session. It seems that in Zoom, a single (or multiple) bad connection won't mess up the whole call.
Also, the quality doesn't seem to degrade as more people are added.