Sometimes Reverse Puzzles are good, sometimes they are not.
Opinions about the same puzzle may differ, so I suggest a pretty simple unbiased criteria.
If there exist a non-hardcoded solution that passes all open tests but fails at least one validator - this Reverse Puzzle is broken and should be fixed or removed.
What do you think?
I agree, in CoC: puzzle where you can fail validator when all tests passes should be reworked
non-hardcoded is a pretty vague tag. If a solution attempts to solve a problem and it clearly has its basic behaviour functional, but the problem is in a validator with an edge case where such an edge case isn’t present in the sample tests; regardless if it’s reverse mode, shortest mode or fastest mode, it should be fixed.