this is quite the nitpick... for all intents and purposes, sha3_256 will _not_ collide; it's infeasible to the point that it can be said that the resulting hash is unique. I mean, of course you are technically right, but sometimes all the details are not worth mentioning when explaining a lot of small parts of a bigger picture.
If you notice, it only exposes the _types_ of the TripleDES, not the constructors, meaning you don't have access to the values unless you clone the cryptonite repo and modify it to expose the constructors, giving access to the values at run time.
Also, the tutorial is a bit advanced and is meant to show a particular use case of using symmetric block ciphers for encryption/decryption, you are not always using such bare-bones primitives-- check out the hashing part of the README.md in Nanocoin.
IMO it's production ready, and has most all potential known attacks documented above the functions that are vulnerable.
It's a bit simpler in implementation; the relevant data structures are defined a bit differently, so it could give a nice alternate perspective about what a blockchain written in Haskell may look like.
News flash: Strong statically typed languages make software engineering easier! ... wait, this is not really "news", is it? Anyway, I'm happy that well known companies are advocating it.