The interviewers clearly expected the candidate to give up on his weird solution early. They should have just said, "Do it another way, we don't like this." But they let him dig his own grave instead. This solution is impractical in many cases, and fundamentally difficult to debug. They said many things that suggested that they wanted "easy to extend" and that implies "easy to debug." I don't really have much sympathy for the candidate. The whole story could also be fake, and certainly sounds like the kind of thing someone would make up to post a clever solution to HN that in truth took days or weeks to come up with.
It seemed like it was actually trivially easy to extend. I’m not saying it’s an optimal solution in general, but given the absurd requirements, I think it was well done.