I must be misunderstanding. Every test run was spending 3 _hours_ just discovering which tests to run 9 times?