Simple ground fault indicators work by measuring voltage from the circuit conductors to ground. As lots of people have already mentioned, if a phase goes to ground then the voltage to ground will be low everywhere on that phase, and all of the indicators on that phase will activate, everywhere in the connected system. (Sure, there will be small differences in voltage, getting closer to zero the closer you get to the actual faulted system...)
To actually find the ground fault you need to look for current flow. big john mentions one technique, intentionally 'pulsing' a supply connection to ground and looking for the pulsing current in your feeders and circuit conductors. I believe that there are other possible techniques, such as injecting higher frequency common mode AC (AC voltage between the system and ground at some frequency different from the mains so that you can detect it.) But in all cases you would need a tool to look for current flow.
In North America, ungrounded and high resistance grounded systems are used for 'continuity of service', to permit a system to operate during a ground fault. It is entirely possible to design a system solely for the purpose of reducing ground fault current, not maintaining operation. This requires the use of ground fault detecting breakers, since ground fault current is low and would not trip OCPD. I don't believe such 'fault current reduction' fit well with NEC requirements, but there doesn't seem to be much in the way of market.
-Jon