Assuming that the author had "pedagogical use cases" in mind, I'll have to ask: why not just a plain server with ssh access and screen w/ multiuser support installed? It's really hard to believe that it would take anyone more than 10 minutes to learn how to use screen to the point they can share a session remotely.
If anyone else is struggling with screen, I recommend launchpad.net/byobu, for a more humane/learnable/menu-based multiplexing program
1) open a terminal
2) ssh joe.student@teacherserver.net
2.a) your password is "PASSWORD_FOR_JOE"
3) screen -x joe.session
Share files with a friend
Run a (sinatra web app? whatever) and get feedback on it for 20 or so minutes
Do anything that requires a few open ports.
It's more than just a shared terminal.
What I don't understand is how one can justify spending all this time to build one application that is an inferior, resource-hogging clone of a GNU tool that is probably older than him. And the only possible explanation that you could give me is "you could run this on ChromeOS"?
I don't mean to be rude, but it looks like you are saying the right things, but not stopping to wonder if it makes sense or not. Is it really that much of an advantage to have this "feature"? If all you have is a ChromeOS device, you are probably not the person who would need to have an Erlang REPL. And if you are that kind of person, you'll probably get a computer that runs an actual operating system, one that actually gives you a terminal and a SSH client.
I know to you and me these aren't major hurdles. But we have years of experience and knowledge of these "basic" tools that most other people don't. So as a teaching tool for those without such experience this is great. Also I think you underestimate how much inertia needs to be overcome for most people to try something new. This drastically reduces the friction to trying out new languages.