A new cooperative fault accommodation algorithm based on a multi-level hierarchical architecture is proposed for satellite formation flying missions. This framework introduces a high-level (HL) supervisor and two recovery modules, namely a low-level fault recovery (LLFR) module and a formation-level fault recovery (FLFR) module. At the LLFR module, a new hybrid and switching framework is proposed for cooperative actuator fault estimation of formation flying satellites in deep space. The formation states are distributed among local detection and estimation filters. Each system mode represents a certain cooperative estimation scheme and communication topology among local estimation filters. The mode transitions represent the reconfiguration of the estimation schemes, where the transitions are governed by information that is provided by the detection filters. It is shown that our proposed hybrid and switching framework confines the effects of unmodeled dynamics, disturbances, and uncertainties to local parameter estimators, thereby preventing the propagation of inaccurate information to other estimation filters. Moreover, at the LLFR module a conventional recovery controller is implemented by using estimates of the fault severities. Due to an imprecise fault estimate and an ineffective recovery controller, the HL supervisor detects violation of the mission error specifications. The FLFR module is then activated to compensate for the performance degradations of the faulty satellite by requiring that the healthy satellites allocate additional resources to remedy the problem. Consequently, fault is cooperatively recovered by our proposed architecture, and the formation flying mission specifications are satisfied. Simulation results confirm the validity and effectiveness of our developed and proposed analytical work.