BTW one of the arguments I have heard against IG is that the "one way" penetrations through metal cabinets and raceways that are not bonded at both ends actually creates a choke that reduces the effectiveness of the IG. (similar to GECs)
If you have a single conductor in a ferromagnetic raceway, then you will get the choke effect similar to that experienced by GECs (and the reason that conduit surrounding GECs must be bonded at _both_ ends.
If the conductor is carrying fault current from the circuit conductors _in the same raceway_, then this 'choke effect' is neutralized; any current flowing in one conductor is balanced by current flowing in the other.
IMHO 'run with the circuit conductors' clearly means _inside_ of the corresponding raceway if the raceway is metallic, for the same reasons that phase conductors must not be run in separate metallic raceways.
Also, if the feeder conductors are run inside of the conduit, and the "quiet ground" run on the outside, then there will be considerable coupling between the "quiet ground" and any currents flowing on the conduit. This would seem to add noise to this quiet ground.