Indeed, having automated tests would be a major improvement and a way to ensure as few regressions as possible.
But there would be so many things to test; look at this example, who would have thought to write a test for the "Alt+F4 while cursor in window" case? Sure, once you get a bug, you write the test that checks it, but generally after you get a bug it's fixed and rarely comes back.
And of course it would be very hard, and require a huge effort. But if you want to try, that's great, thanks a lot