My experience is that the tests should either test functions that are small and do one thing (I.E. sorts, maps with some logic, basically where you want to test

My experience is that the tests should either test functions that are small and do one thing (I.E. sorts, maps with some logic, basically where you want to test edge cases and sanity check). In those cases there is very little reason to change that code. If you are testing something larger, the test should be an integration test, where you test the full business logic flow. That makes the code less PITA to change while still giving you confidence. If the business logic actually changes, the tests should break IMO because they are there to ensure that the business logic remains consistent. When you test the business logic (without testing the implementation) the code becomes much safer to modify and refactor.