Time-Reversal multiple signal classification (TR-MUSIC) is here applied to testing cable networks in order to detect and locate soft faults. TR-MUSIC is shown to provide spatial resolution in the millimeter range while using continuous-wave (CW) test signals, even at frequencies with guided wavelengths much larger than cables length. State of the art time-domain reflectometry (TDR) methods would require bandwidths in the order of hundreds of MHz for a similar performance. As opposed to TDR, TR-MUSIC does not suffer from the ambiguity created by the existence of multiple echoes in cable networks, which can be easily misinterpreted as multiple faults, leading to false alarms. TR-MUSIC is intrinsically adapted to dealing with multiple faults, handing a direct estimate of the number of faults found in a network under test. Furthermore, the detection capabilities of TR-MUSIC are insensitive to the severity of a fault, as faults are not detected based on the intensity of their echoes as done in TDR techniques, but on a sub-space approach mostly dependent on phase patterns. Accurate identification of faults from CW signals points to the possibility of designing simpler test systems, not requiring pulse generators and fast electronics. TR-MUSIC accuracy is demonstrated experimentally for locating both single as well as multiple soft faults in two cable networks. The proposed method also gives access to the reflection coefficient of each fault, thus enabling an estimate of its severity.