One thing that might be neat in the future is noticing differing results
while the tests are still running, and modifying the log messages to
indicate them. I can imagine situations where you would want to abort
the test run immediately after seeing the first regression.