If your refactoring causes test code to break, then it also causes "real" code to break, yes? All code using that interface needs to be fixed appropriately, whi

If your refactoring causes test code to break, then it also causes "real" code to break, yes? All code using that interface needs to be fixed appropriately, which can typically be done with automated tools. If there's no corresponding production code using the same interface, you have a problem: you're probably testing at the wrong level.