Interesting to see you can jam something like this into Haskell despite its inappropriateness. The (genuine) Agda proofs are much, much nicer.

Agreed :-)

One of the big questions to me now is whether we can prove the Diagonal Lemma, not just assume it. Do you know if it's possible in Agda? I heard somewhere that it might be hard to find an intuitionistic proof.