I think there's an issue with how our logs.py sets up logging.

It has a start and stop function, using a global variable to prevent
multiple starts without a stop in between. The problem is, the stop
doesn't properly clean up the configuration that's done in the start
method. So each time it's stopped/started, we add all the handlers back
onto the loggers.

This shouldn't be a problem (currently) in the running code since it's
only started in the application bootstrap. However our unit tests are
badly broken. I noticed in the debugger that there were a ton of
duplicate handlers on loggers when running tests. The stop/start process
is tied to each time the configuration is loaded in a test, which is
done in setUp. Since that's called once for each test, it's causing our
logging configuration to blow up by the end of the test runs. As of
right now, the test_consumer_history_api tests hang for both me and John
Matthews. They run fine when I comment out the testutil.start_logging()
call.

This might also explain why Hudson is hung right now (current build has
been going for 15h). Python seems to get especially confused in the
auditing logging configuration since it's adding a bunch of
TimedRotatingFileHandlers. I'm curious if it

I'll file a high priority bug for it since it's stopping our tests and
builds from running. I'm not 100% sure of the fix, but the issue looks
to be that our stop/start logging calls are bugged.