Mutation testing answers: do the tests actually catch bugs, or do they just execute code? The 96% coverage metric tells you how much code the tests reach, not how well they catch errors.

Run mutmut against critical modules first:

pip install mutmut
mutmut run --paths-to-mutate=vera/checker/ --tests-dir=tests/
mutmut results

Start in detection-only mode (don't enforce kill rates). Focus on the type checker and verifier modules where a surviving mutant indicates a silent correctness bug.