A method for tracking agents across loss of state events is described. After determining the number of terminal agents within a hierarchical agent system, an algorithm forms a first matrix containing data identifying agents within the hierarchical agent system. After a potential loss of state event has occurred, the algorithm forms a second matrix containing data identifying agents within the hierarchical agent system and compares the first matrix to the second matrix. If the matrices are identical, no agent switch occurred during the potential loss of state event. If the matrices are not identical, at least one agent switch occurred during the potential loss of state event.