They could always work on making the system peer-to-peer. You will never get 100% of the way there because of some overly restrictive networks, but you can get the majority of users (especially those on home connections without enterprise firewalls/etc) to connect directly making it sustainable enough (the last 10% can just be subsidized by paying users and/or by ads - it could literally be a "fix your network for P2P to work, or watch ads to pay for TURN relays).
P2P multiparty videoconferencing just isn't realistic outside of tech demos. In a call of 20 participants, someone with a typical 720p webcam would be transmitting more than 50Mbit/s, which doesn't sound like a lot when ISPs market so-called "1Gbit fibre", but in fact is more than a lot of residential connections and home Wi-Fi can reliably push without unacceptable loss/delay/jitter.
A product built with this architecture could not deliver the reliability and performance that Zoom delivers with the variety of connectivity quality that its users have.
consumer routers are not the main problem with that (they might even support multicast, because various ISPs use it to distribute TV channels) - the public internet generally has no way of handling it.
WebRTC can support it all they want, but to a first approximation, Multicast doesn't work over the open internet, so it won't help outside of corporate videoconference (which generally runs OK, because corporate networks are mostly wired and often overprovisioned or easier to provision than homes)
