This paper elaborates about the potential risks of systemic instabilities in future networks and proposes a methodology to control and mitigate them. The starting concept is modeling the network as a complex environment (e.g. ecosystem) of resources and associated controllers in a continuous and dynamic game of cooperation and competition. Key observation is that Internet might be viewed as "self-organizing" and that this is achieved through "constrained optimizations" (through protocols methods, algorithms, etc). The methodology foresees defining and associating utility functions to these controllers and elaborating a global utility function (as an aggregated function of local utilities) for the overall network. It is conjectured that the optimization of proper global utility functions ensures network optimization and stability at the same time.