So knowing all this, what I was planning to do was to debug and understand what was the steps that caused those processes to stay open, using the test as a trigger and verifying that it's 100% sure that the test is not the cause. Then if it wasn't (so a bug is present), fix that bug; then at least try the common shutdown cases (shell, service, especially on Windows). If that work I think it would be good enough vs before that PR. Knowing that the test is not 100% reliable I was just using it to trigger unexpected situations.