This is partially a flaw and partially a feature. It should be common sense that joining a massive room on a tiny server may overload the server: it has to talk

This is partially a flaw and partially a feature. It should be common sense that joining a massive room on a tiny server may overload the server: it has to talk to all other servers in the room (which is where most of the resources go), and keep track of the current state (ie key/value pairs such as who’s in the room). It’s also a flaw in that this could be less heavy: it could be faster to join (which we’ve almost solved via Faster Joins - see https://matrix.org/blog/2022/12/25/the-matrix-holiday-update-2022#turbocharging-matrix https://matrix.org/blog/2022/12/25/the-matrix-holiday-update...). We could also speak to a subset of the servers in the room (at the expense of reducing the decentralisation of the room). We can also optimise state resolution further. For now: if you want to participate in a room whose users are spread over 1000 servers, you’ll need a server beefy enough to talk to 1000 servers every time someone says something.