Internet access from space enjoys renaissance as satellites in Mega-Constellations is no longer fictitious. Network capacity, subject to power and computational complexity constraints among other challenges, is a major goal in this type of networks. This work studies Network Coding in the presence of dynamically changing network conditions. The notion of generalized acyclic network is introduced and employed for promoting the generation of linear-multicast network code for what is considered to be a cyclic network. The performance of several network coding schemes, among these is the known static network code, is evaluated by a STK simulation for a swarm of communicating satellites, conceptually based on the Iridium system. Exploiting the prior knowledge of the networks topology over time, new network coding approaches are described, whose aim is to better cope with the time-varying, dynamic behavior of the network. It is demonstrated that in all cases, pertaining to our example network, static network codes under-perform compared to the presented approach. In addition, an efficient test for identifying the most appropriate coding approach is presented.