I'd say a good negative use case really fits in the "reliability" category. So often at work, coders expect everything to always succeed, and have no thought towards what happens if one cog ever falls out of place; but good systems can react well or even help you get to what you generally need.