You can build whatever interaction system you want on top of the thing. Calla is a wrapper around Jitsi to add the spatialization and communicate user state between users.
As for the server costs, I'm not going to use anything other than Azure. Most of the cost is bandwidth, but it's worth more to me to keep everything in one place and not have to worry about learning another PaaS system. From the other offerings I've seen, Azure is usually only more expensive at the most lowest of tiers. Once you start scaling up, the other vendors get just as expensive. So I'm fine with shelling out a few more bucks here and there to avoid having to use something else.
https://github.com/capnmidnight/Calla/blob/master/Calla/doc/...
As for the server costs, I'm not going to use anything other than Azure. Most of the cost is bandwidth, but it's worth more to me to keep everything in one place and not have to worry about learning another PaaS system. From the other offerings I've seen, Azure is usually only more expensive at the most lowest of tiers. Once you start scaling up, the other vendors get just as expensive. So I'm fine with shelling out a few more bucks here and there to avoid having to use something else.