I think I obtain a lot of the benefits of focus without quite being this extreme. I use a lot of virtual desktops. One virtual desktop per logical activity. For example as a student it would be one desktop per course I'm studying. As a software engineer, each git-repo that I touch gets its own virtual desktop (especially when one application involves many git repos). Each virtual desktop has its own editor process, terminal, and browser window for documentation and tabs related to that activity only. This means I can change my focus just by switching to another virtual desktop. And the tabs and editors and such on other desktops are not visible (they are not even in the panel) while I'm not working on them. So unlike you, I have lots of bystander applications, and each one is exactly where I left it. They just aren't visible until I switch to that activity. So I would argue that they don't distract me at all.

I also have things set up so that there is a keystroke that raises the editor on the current desktop to the top. Another for the browser and another for the terminal. When I switch to another desktop, the same keystroke raises a different editor process for that desktop, etc. So within a desktop, I switch between these apps using the same keystrokes regardless of which desktop I am on. And there is only one way to switch desktop. I can't switch to another desktop via choosing another application on another desktop, because the other applications that are not on the current desktop are completely invisible.

When I'm working on something with a Scala backend, an Elm frontend, and C/C++ embedded device, and when I need to go back and forth often between these sub-applications making related changes, I can't afford to close and reopen each one when I change the language/repo. I can change desktops very often this way. eg. Add a new field to the frontend. Immediately add it to the backend and the embedded also. But they are in different repos. etc. etc.

If I have one editor opening all three projects it is a nightmare to navigate between files. But one editor for each language and repo is the sweet spot for me. And it means my documentation for Elm is all in a separate browser window from my documentation for Scala (which is on another desktop) etc etc.

Sorry this was so long, but I think all the details add up to making it an efficient system.