[robryk]

My best example of that was a test that was asserting that a monitoring metric changes in some way with a comment expressing doubt whether the asserted values are correct (but whoever wrote it still wrote the test that enshrined the behaviour that they themselves doubted).