On Windows, the proper thing would probably to make sure we spawn child processes in a job and open it from rdctl when we do the factory reset to terminate the job. However, it's unclear how we'd assign the processes into the jobs from Node; calling Win32 directly seems like it would be annoying (since that means doing native bits again and it was painful last time).

Presumably on Linux/darwin we'd do something with process groups, though it's unclear to me at this point what the process group leader would be.

Summarizing the state of things:

• On Windows, we use jobs to ensure that all processes spawned by extensions are terminated on application exit (Win32: Spawn extension processes in job #7838, rdctl: Fix obvious factory reset errors. #7953)
• On macOS, we use process groups to do the same, though that is not as reliable as subprocesses can make their own process groups (*nix: Kill process group when main process exits #7846, rdctl: Fix obvious factory reset errors. #7953)
• On Linux, we attempted to use process groups but have that backed out at the moment because when installed via rpm/deb the electron process does not get its own process group by default (*nix: Kill process group when main process exits #7846, rdctl: Fix obvious factory reset errors. #7953, rdctl: Don't use process groups on Linux #7965).

Therefore the work left is to fix Linux to make it work again.