You don't. Not without further investigation. You need to figure out why they are different.

Sometimes this technique catches bugs on prod that are unknowingly fixed with refactoring.

The test is done so you verify that your new code has not any side effect that had changed previous behavior.

How do you know that your tests are correct? The problem is equivalent.