I am just curious.

Suppose some problem has a randomized solution which fails with probability at most . If it has, say, 100 tests, the probability of failing some of them is about . Therefore, once in 10⁴ it may happen that correct randomized solution doesn't pass.

Has anyone experienced anything like that? Maybe even the exact solution that didn't pass got AC after submitting it the second time? Or am I getting something wrong?