A method is described for determining, via a local fault poller, a list of local unreachable devices of a plurality of devices within a local domain, wherein the local unreachable devices are coupled for communication with the local fault poller. The method comprises determining selected neighboring domains of a plurality of neighboring domains, wherein the selected neighboring domains comprise neighboring devices which are coupled for communication with the local unreachable devices, each neighboring domain having a respective neighboring global fault poller. The method comprises requesting an operational test of the local unreachable devices from the respective neighboring global fault pollers of the selected neighboring domains, and receiving responses, via the respective neighboring global fault pollers, indicating whether the local unreachable devices are operational. The method further comprises determining, based on the responses, a root cause device from the list of local unreachable devices.