In this paper, we investigate the performance of the dynamic allocation of resources between separate cellular networks. We propose a Dynamic Internetworking Carrier Aggregation (DI-CA) framework which involves every network operator releasing some of its exclusive, but excess, spectrum to another network operator for a limited time. We derive the basic condition for which DI-CA can improve the performance for all the operators and then propose a distributed scheduling framework that uses coalition formation with uncertainty, in which each independent operator can decide whether or not DI-CA can improve its performance without having information regarding channel conditions or load experienced by other operators. We propose a distributed Bayesian coalition formation algorithm to approach a neighborhood of the Bayesian Nash equilibrium.