First, it's worth noting that the output of the mutation test is also an heuristic. Secondly, I have really hard time imagining a case where, a change in branc

First, it's worth noting that the output of the mutation test is also an heuristic. Secondly, I have really hard time imagining a case where, a change in branching logic in code, that doesn't trigger test failure, is not signalling a "coverage smell". Edit: And I do totally relate to what you are saying about change traps. And, I think, if such practice is already prevalent in the codebase, mutation testing will exacerbate it.