Makes you wonder why no one thought of this before.
Well, at least one person did, the developer of DiceStream who I stole the idea from. :) But I think the reason it's not a more popular approach has to do with the software tools that are available and the way people tend to think about software. Getting a live video feed is already a somewhat nontrivial task, so that tends to be either its own standalone app or a library that does it. If they don't provide hooks that allow you to composite other stuff with it then it's hard to do with a standalone app, and if it's a library then you've got "network effect" problems convincing people to use your virtual tabletop app that has integrated video (plus all the cross-compatibility headaches). Or if you're looking at it from the "game engine" side of it, your mind will tend to end up with map or environment simulators like roll20 or tabletop simulator because that's the kind of things those tools usually seem good at creating. Plus there's a much bigger market for tools that support those kinds of games.