I found it strange that Martin suggests that the gaps in TDD-style tests with regard to dealing with bad input should be filled by integration tests, which due

I found it strange that Martin suggests that the gaps in TDD-style tests with regard to dealing with bad input should be filled by integration tests, which due to their complexity seem more likely to cover the happy path, rather than fuzz testing or quickcheck (https://en.wikipedia.org/wiki/QuickCheck https://en.wikipedia.org/wiki/QuickCheck) style tests.