Break Your Code on Purpose: A Kill-Rate Gate for Characterization Tests

A green characterization suite proves very little. It proves the tests ran once against current...