I think we can. Following all the implications down the rabbit hole, you end up with a system architecture where there's no front-end, and there's no back-end.
Since CRDTs are still kinda new for most people, the discussion hasn't really gotten there yet.
What is a good starting point to follow all the implications down the rabbit hole?
And wrangling CRDTs into a client-server architecture is actually extra work. The basic design assumes you can trust every peer -- making sure that jes5199 actually was the peer who added a comment that is marked as author: "jes5199" is not trivial.