Regarding the point about a test conveying little information if it never fails: That's true if it has never failed in the past and will never fail in the futur

Regarding the point about a test conveying little information if it never fails: That's true if it has never failed in the past and will never fail in the future. E.g. maybe it really is tautological or it's testing legacy code that has been locked down forever. But if it does fail at some point in the future after having passed in dozens of iterations, that failure will convey a lot of information. Every time it passes, the information-theoretic value of that future failure increases. Unless running the test involves some onerous cost, there is no reason to delete it. Just hide it in a folder where you won't have to look at it.