@DanielLSM Ah, sorry, I understand. In the worked example, they wording "they are the same" is really just short for "they both have their individual DAGs which happen to look exactly the same". There is one DAG per service_intention, not per problem instance.
Yes, you may collect all individual graph for the service intentions into one big DAG for the problem instance, such that each route is a subgraph of the big DAG. However, our routing_graph utility script does not do that (you would then need to encode also the service_intention.id into the ID of the graph nodes, so that nodes belonging to different service intentions can be distinguished).