I think what this proves is that the LLM knows the riddle, and is trying to give the expected answer without paying attention to the insane wording. So maybe this is a good way to fool an LLM.