It is definitely true that pagination could be much easier. The way it currently works however seems fine to me: You query a node with say the rate limit of the first 30 children. Then you store the key of the last result, and use it as the starting point for the next limited query. This way you can easily query a node with 1000+ children without overloading the client.