Cockroach nodes were installed in the following:
1. Single host, triple nodes (containers): for evaluation on a low-end PC.
2. Single host, triple nodes (process): to test applications against a ~500GB database in-house.
3. Serverless: hosted in Google Cloud Platform (the main database).
A number of Python scripts and some Java applications are happily reading and writing to the database.
The solution allows for scaling in cases where a PostgreSQL server (unless you use sophisticated partitioning across some machines) would not be enough to handle the load. This kind of database is particularly used for backtests.
I am a freelancer. A client of mine wanted a solution that would allow them to scale yet not abandon the familiar PostgreSQL front-end (and rewrite a part of their source code).
Scalability aside, CockroachDB is a fine way forward from PostgreSQL and is not changing the client source code part of the system. If you are lucky and you do not use newer features from recent versions of PostgreSQL or PostgreSQL extensions, it's fine.
There are nice-to-have features for big organizations like regional tables. At the moment, my client simply does not use these. However, the serverless offer from CockroachDB is reacting well as data grows.