> To reproduce the test_logging cleanup problem, insert skipTest() in setUp()

My understanding is that skipTest would normally be raised in the test method itself or as a decorator to it, and not in setUp() itself. (It wouldn't make sense to, as that would skip every test in the test case - not the obvious thing to do.) I can understand that failures that happen in setUp() may cause tearDown() not to be called.

I would guess that setUp() should recover from any problem and set a flag which is then used to skip in the other tests (which rely on that problem not being there).