The use of anycast routing in challenged networks, which include space networks, can help to alleviate the performance impact of the intermittent network connectivity and large communication delays that characterize such operational environments. With the anycast service, each data bundle can be delivered to any member of the anycast group and by establishing an adequate routing mechanism, bundles can be directed toward the member offering the lowest end-to-end delivery time. This possibility is explored in this study through a design extension to the Cognitive Space Gateway (CSG) approach to DTN routing. The advantage of the proposed method is that it adaptively determines the path for each bundle considering the estimated network congestion, which allows dynamically deciding the (near) optimal member as the destination node. The process allows for low bundle response times and high throughput. Testbed experiments are reported which confirm the performance advantages of the proposed method over unicast routing.