in api_test/middleware/api_test.py#L31 the code is written to execute the custom tearDown method first before it runs the custom setUp method. This ensures that only one method must be called to start a test but exposes a problem where a user could create an object in the setUp method they want to destroy in the tearDown method. Since the tearDown method gets called first, on the first run of the code it would raise an exception.

A possible solution would be to have some kind of "dirty" flag that indicates whether the server was just recently restarted or that the custom setUp method was run at least one time. Then only execute the tearDown method when the dirty flag is set.