We propose a network of excitable systems that spontaneously initiates and completes loop searching against the removal and attachment of connection links. Network nodes are excitable systems of the FitzHugh-Nagumo type that have three equilibrium states depending on input from other nodes. The attractors of this network are stationary solutions that form loops, except in the case of an acyclic network. Thus, the system is regarded as a loop searching system. To design a system capable of self-recovery (the ability to find a loop when one of the connections in an existing loop is suddenly removed), we have investigated regulatory rules for the interaction between nodes and have used two characteristic properties of nonlinear dynamical systems to provide a solution: postinhibitory rebound phenomena and saddle-node bifurcation.