High-volume synchrophasor data generated by the wide area measurement system (WAMS) present new challenges for the design of scalable and robust communication architectures suitable for the large-scale deployment in power grid monitoring and control applications. Features of information and communication technology (ICT) can be harnessed to devise a more flexible infrastructure than the conventional unicast-based hierarchical architecture. Multicast has been suggested as a potential solution for scalability issues in the current WAMS architecture. This article proposes a decentralized dynamic multicast routing management algorithm that manages network traffic efficiently, improves scalability, and makes the system resilient to component failures. The algorithm models the creation of alternate links as a stochastic process using the random-graph theory and manages connections between end hosts at the session layer. This protocol can be implemented using existing ICT protocols at a minimal cost. Numerical simulations and study of frequency control in an IEEE 39-bus system testbed show that the proposed algorithm improves the topological efficiency and resilience of the network.