Again, this is not a universal truth. A lot of brand-new code is entirely a new code path. Old unit tests simply won't exercise that code one way or the other.

Or take the case of fixing a production bug. Clearly no unit test covered that code path in the first place. To properly fix it, you need to write a new test which detects this issue, but none of your old tests will fail.

I spend quite a lot of my time optimizing or refactoring. The last thing you want to see after either of these is code breakage, because it means you made a mistake.