Many functional tests don't call stop_servers after start_servers

I see many tests in the functional tests that don't call stop_servers at the end, after a call to start_servers at the beginning. test_index_with_https_auth in test_bin_glance.py is one such example, but more examples can be found in test_bin_glance.py (and maybe other files).

I don't know if this is intended or not (clearly, not having to restart the servers is a speed win). However, one net result I can see here is that I do have some processes from the test around after the tests have been run. And that is bad.

So should all tests calling start_servers() also call stop_servers()? Or should we have a way to automatically stop the servers?