The way that the bucket resources are implemented is incredibly restrictive.
Currently marked as an improvement, but I was close to moving to a bug considering this is a regression in behavior.
For example, it's now impossible to have 2 clusters in a single namespace with the same bucket name but with different settings.
This would be a problem in dev environments as it's often non-trivial to change bucket names within an application due to it being hardcoded in n1ql queries, connection code etc.
A second problem is that Kubernetes metadata names have a fixed regex that does not match the regex for Couchbase buckets.
An example here is the use of underscore, which is very popular for use in bucket names.
Moving from 1.2 to 2.0 it's now impossible to use your old buckets which had underscores in.
Both of these issues would be solved if there was the ability to decouple the bucket name from the name of the Kubernetes resource.