and is silently skipping when a specified usemodules module isn't found really a good failure mode?
also, more generally, i find skips are pretty loosely applied in tests, seems like it could lead to many silent skips which open up bugs. ie if not hasattr (mod, func): skip (fork doesnt exist on windows or whatever) rather than testing/skipping based on platform directly (doesn't fail, would just start skipping if os.fork happened to disappear on linux). fine for app level code, but really seems like too much inference for a test suite. do you think it's worth cleaning things like this up?