[_delirium]

An even better (and common) strategy is to reject the work that fails the gold-data test and to reject most of the rest of the work too, claiming it failed the gold-data test.