As mentioned on the pull request and here, the fix for incorporating the AttributeConverter calls into the Type system ended up being very simple fix. Look at the 'intermediateJavaTypeDescriptor' reference in (newly refactored out) org.hibernate.mapping.SimpleValue#buildAttributeConverterTypeAdapter method. I also beefed up the comments in that bit of code to hopefully make it more clear what is happening and why.

org.hibernate.type.descriptor.java.JavaTypeDescriptorRegistry may still have some design concerns. Its a new concept and the best approach for it is still being felt out. I did end up having calls that attempt to resolve the JavaTypeDescriptor for Serializable impls flow throw a different logic branch. I think ultimately this should be limited to "standard" JavaTypeDescriptors, similar to the plans for org.hibernate.type.descriptor.sql.JdbcTypeJavaClassMappings.

I am leaving this open for the moment as I want to clean up some of the code comments and am running this through the JPA TCK to see what else it might shake out.