There is increasing awareness of CRDTs but it’s still unclear to many developers how to integrate them with sql. It seems like a CRDT requires developing a new app, or at least a new data model, rather than benefitting from their existing knowledge.

This should change over time; the libraries and documentation keep getting better.

I recommend Automerge 3.0’s documentation as an introduction to the concepts and how to bridge with traditional data. https://automerge.org/docs/hello/

I think the biggest challenge is brownfield.

In greenfield it's easier (so long as you have the grit to build the base pattern); Events get put into tables (TBH this can be optional, but if you're new to it just do it b/c it will make debugging easier,) and if you want a friendly clean 'view' state you have some form of map-reduce replication running on the event set.

And yeah you can even remix that idea to work on 'existing' tables but it's gonna be more work and you'll probably want some form of optimistic concurrency check...