In the real world, I believe a human could figure it out based on the permutations that make up a more or less valid message compared to the others. For example, if my cat learned Morse code but wasn't great at timing, the message "FEED ME" would be more likely than "FE ED ME" due to its simplicity. However, the code itself cannot understand the context. And my cat is almost 18, cut him some slack.

If we were guaranteed that all words belonged to a provided dictionary, it might be possible to solve this issue. However, as far as I can tell, we're stuck hoping that we have made the same assumptions as the test authors. Alternatively, we could dump their tests into exception messages to see what they're feeding and start special-handling the ambiguity as we go. It's impossible to solve an ambiguous problem unambiguously, and I'm not seeing that this one is fully defined.

Maybe that's just me and I'm just not clever enough, which is a very real possibility too.

I agree!!