If anything, I think videoconferencing really needs to move towards more abstracted forms of interaction instead of trying to accomplish crystal clear video. Evidently, remote classrooms are finding it easier to do lectures via Minecraft than Zoom or Skype.

With the kinds of advances in deepfakes and realistic looking face rendering we're getting, there isn't as much reason to be sending that much data over the network. You'd need a camera to capture things like writing on a markerboard, but just for conferencing and getting body language pseudo-MoCap would do fine.