There are at least three issues to resolve / take decision on for EIP-712 in my opinion:
- array hashing (as you mentioned)
- the origin proposal (including the non-interactive signature bit) : see comments on EIP-712 and my blog post
Note that while in the latter comments from me and @Recmo a solution involving a special envelope is mentioned. I actually disagree with it since it would force standards built on top of EIP-712 to choose whether they support origin or not. Origin should be part of the raw EIP-712 proposal like chainId is - chainId and blockNumber pair for replay protection on minority-led forks => if blockNumber is included, they can’t be part of a static domain separator as blockNumber would be updated frequently, see discussion on EIP-1344, EIP-1959 and EIP-1965