It's really freeing when coupled with something like Gitpod or CodeSpaces. You join a new project and you have ZERO setup to do on your own. You can open up a new branch and start working in seconds. I don't have to worry about whether my computer has enough memory to keep up with Docker, if it's fast enough for this new project, if I have to install version managers to separate runtimes for each project (rbenv, pyenv, nvm, etc).
To perform experiments it's even better. I have multiple directories with sample/initial projects from many languages where I was just feeling out a new framework or library or concept. Then months pass and I can't run them again because the specific dependency is no longer installed on my laptop. Now I can create these new study projects and know that it'll run in seconds when I get back to it.
Plus not all of us can work from highly performant workplace-provided workstations. I have just my laptop to work with. Imagine running Docker and Chrome and Spotify and maybe macOS decided that was the right time to hog my disk and do file indexing while I'm running tests...it's a hassle I'm glad to be rid of. I'd rather a company get me a Gitpod/Codespace and I can separate things cleanly.
My first experience with such setup was the DG/UX at the university campus, where we had a mix of classical green/ambar phosphor terminals, and a couple of X-Windows thin clients from IBM to connect to it.
Regarding Multics, it was a failure only for Bell Labs themselves, as the project kept going and had still a couple of years for itself.
That sounds horrible.