What type of backend do you need in terms of servers/users if this was a real product? Clubhouse has been suffering under heavy load in certain cases. WebRTC is mostly P2P, but there is still a significant server load to handle edge cases and large rooms.
You should show this off to Mozilla or some other web company that's backing WebRTC and try to get some funds for scaling up your servers if you need to in the future.
We also want to make it super simple to spin up your own Jam for example if you are organizing a micro conference over a weekend and for that we will need to better understand what that means so we can recommend what kind of server/specs to go for.
Thanks a lot for bringing this question up. We need a good answer.
We are currently reaching out to friends and strangers who have a bit more experience running high demand WebRTC + mixing setups to learn about the scaling challenges.
That said, we are quite surprised how well p2p WebRTC works for audio-only in smaller rooms if most participants have a decent internet connection.
WebRTC + Opus is magic.
If/when you hit scaling challenges I would love to help! I maintain github.com/pion/turn and github.com/pion/webrtc. One of the reasons I built it was so that I could put my TURN and Signaling server in the same process. It makes it way easier to tie your auth together for signaling+TURN. Then if you do go down the SFU route lots of interesting things you could do. You can see that with how screego[0] does it.
Happy to help however I can (even if not using Pion!)
[0] https://github.com/screego/server/blob/e845b3d29c4b5794ed10f...