[Reefersleep]

Turn it upside down. Do the bare minimum of thinking and writing when setting up the initial test case, really just have it be scaffolding to run a single sunny scenario test case. Then let users (testers, other developers, automated systems or actual users - whatever hits your code) do any further discovery for you. Whenever they cause a bug, that's a new test case (or a series of new cases) to add to your initial single case.