[wpietri]

That's definitely not true. Tests that lock in a bug or an accidental epiphenomenon of code are demonstrably worse, because they actively discourage improvement.