We stumbled on a problem where we had to compare state-based with behavior-based testing in order to get our context and mocks right. Using Kent Beck's recent composability article as reference.