is there a reason an acceptance test case has different behaviour when run by itself vs when run as part of the test suite? seems like the various test cases, though ordered, do not rely on each other