this is rather standard behaviour for test suites: setup and teardown before and after each test. The philosophy is that tests should not depend on each other. If you want another behaviour, you should probably use static objects that persist between each test.