In a method and system for recovering a faulty unit, a predetermined number of recovery periods, and the maximum number of times a recovery act is carried out during a recovery period, are defined. A faulty unit is taken out of operation, recovered (i.e. subjected to a procedure designed for remediating the fault, if possible), and brought back into operation if tests indicate that the unit is in condition (i.e., is not faulty). The recovered unit is monitored for the time of succeeding recovery periods and is re-recovered, if necessary, provided that the number of the recovery acts carried out during each recovery period does not exceed the maximum number predetermined for the respective period. As a result, bouncing of the unit back and forth in an uncontrolled manner, if the fault is of such a type that the recovery system cannot detect it on the basis of the tests carried out, is reduced and eventually curtailed.