This is definitely not a good idea for some constructive problems, where knowing the jury answer could allow participants to reverse-engineer the solution from the outputs. This is a bit less of an issue for problems with fixed answers, but something like that can still happen to a lesser degree.

You can easily know you've misunderstood the problem when you receive WA verdict.

Don't turn Codeforces into HackerEarth.