Rambo is the Reconfigurable Atomic Memory for Basic Objects, a formally specified algorithm that implements atomic read/write shared memory in dynamic, rapidly changing networking environments. Rambo is particularly apt at dealing with volatile environments such as mobile networks. To maintain availability and consistency, even as hosts join, leave, and fail, Rambo replicates objects and uses reconfigurable quorum systems. As the system dynamically changes, Rambo installs new quorum configurations. This paper addresses the reconfiguration problem with three approaches based on a finite-domain model, an hybrid master-slave decomposition and a parallel composite to find optimal or near-optimal configurations. Current behaviors of Rambo participants are observed, gossiped, and used as predictors for future behaviors, with the goal of finding quorum configurations that minimize read and write operation delays without affecting correctness and fault-tolerance properties of the system.