In addition to that, I've got to say this:
IDENTITY columns to me are really 'just' syntactic sugar.
What I like is of course that the table DDL clearly represents the behavior of the column and to a certain degree the intended use (it's still required to define the primary key, though, just as you did).
What I don't like is the fact that the column literally disappears in the code where records are created.
I'd argue that the table definition probably is read less often then the CREATE RECORD code and it would be good to know where those ID values come from.
So, while I think IDENTITY can be convenient shortcut, I don't believe that it makes all that much more work to simply use a custom sequence.
In my eyes the best option here would be to extend the current IDENTITY column model with the option to use a custom sequence. That way people who like to not have to put in <seq>.nextval into their CREATE RECORD code can still use the automatism but also get the option to administer the number dispenser aka sequence as they like.