Delay or Disruption Tolerant Networking (DTN) is a communication paradigm designed to provide internet-like services in networks with intermittent connectivity, particularly in satellite and deep-space communication scenarios. In these challenging environments, traditional routing protocols often fail due to high latency and frequent link disruptions. This paper introduces an extension to the Cognitive Space Gateway (CSG) approach, which uses the low-power, parallel processing capabilities of spiking neural networks to optimize end-to-end bundle delivery. The proposed method enhances routing by adding awareness of storage limitations at DTN nodes along the bundle’s path. The approach is evaluated using a simulated satellite constellation scenario. Results show that the enhanced CSG effectively learns to avoid congestion areas within the constellation, preventing storage overflows. This leads to substantial improvements in throughput and end-to-end delivery times compared to standard DTN routing methods.