Techniques for detecting loops in routes that cross route information boundaries include receiving a control message at a first edge node on one side of the boundary that is connected to a different second edge node on another side of the boundary. The control message indicates a particular network address of a particular node that is reachable from the first edge node. Distinguisher data is determined that indicates if a node in the first collection can reach the first edge node without leaving the first collection. An advertising message is sent from the first edge node to the second edge node that includes route data that indicates the particular network address and the distinguisher data. Based on the distinguisher data, a testing edge node in the first collection can determine whether there is a loop comprising both an internal path and an external path to the first edge node.