#1 is not a real problem on practice. If the concern is payload size, gzip solves it for the most part. If the concern is consistency (e.g. there's an update to

#1 is not a real problem on practice. If the concern is payload size, gzip solves it for the most part. If the concern is consistency (e.g. there's an update to author and it needs to propagate to all books from this author), most graphql clients + view frameworks already handle it by maintaining an id based cache and notifying all users of the said object when something changes.