otherwise we're not actually testing what we want to test, and it's darn hard to track down bugs in the test code
