If the goal is to "cover all code—all branches", then that's testing from a certain point of view, what you might think of as "characterization" testing.
Super valuable, but if the code is actually wrong, and you want to ensure that it does what it _should_ do, that's a different goal.