Since my last update, I have been figuring out how to construct an access list root from a meta-witness. I realised that there were a number of gaps in my understanding around what is currently implemented in wit/0, how that is different to a full Merkle proof witness, and what we would need from a witness in order to construct a verifiable access list root. I spent some time researching these topics and published my conclusions in a review article on witnesses - it can be found here.
Now that I am conceptually clear on how to construct the access list root from the witness, I am working on a python reference implementation of this. I’ve started a new repo here that will host the witness-related analysis code produced during this project. By the next update, I hope to have this implementation completed and proven for meta-witnesses.