The field of Swarm Intelligence is increasingly being seen as providing a framework for solving a wide range of large-scale, distributed, complex problems. Of particular interest are architectures and methodologies that address organization and coordination of a large number of relatively simple agents distributed across the system in a way that encourages some desirable global emergent behavior. This paper describes a SWARM simulation of a distributed approach to fault mitigation within a large-scale data acquisition system for BTeV, a particle acceleratorbased High Energy Physics experiment currently under development at Fermi National Accelerator Laboratory. Incoming data is expected to arrive at a rate of over 1 terabyte every second, distributed across 2500 digital signal processors. Simulation results show how lightweight polymorphic agents embedded within the individual processors use game theory to adapt roles based on the changing needs of the environment. SWARM architecture and implementation methodologies are detailed.