I think what Don is trying to say is that the circuits from the sub station to the local pole mounted transformer that feeds the building can be a problem for stray voltages because these are wye circuits, most substations are fed delta and do not cause stray voltages because there is no MGN that shares a current path with Earth.
I agree with Don as to the problem of stray currents being caused by primary side neutrals, and this is a big problem and will get worse as many POCO's get away from local delta circuits feeding local service transformers.
Using the Earth as part of the path back to a substation lowers the cost by allowing a smaller neutral conductor to be used, but at the same time it is increasing the stray voltage problem when to much current on the MGN is being applied or when there is a break in the continuity of the MGN that causes Earth to be the only conductor.
As far as transformers go they are the start of a new source each time we cross one in a circuit, so any neutral current after a transformer only applies a line to line load ahead of the transformer.
A smaller neutral conductor can be used in either the NEC or the NESC. There are sizing requirements, and the neutral is used to carry imbalance currents. I agree that stray voltage can be a problem with the MGN configuration we use. (ladder ground, picture a ladder turned sideways.) BUT, the interconnection of every other device for safety sake and maintanience on the customer side can also contribute to stray voltage.
If the connection to neutral is lost at the customers meterbase, the currents will seek a return to source through any means necessary(interonnected phone, cable, water, etc.)
SWER systems are the only system I can think of that use earth to be the only conductor. I cannot find in the NESC where SWER systems are allowed. I am still looking.
POCO's use the neutral as a low Z path to source, not the earth. The earth becomes the return path only when this wire is stolen, broken, or connections are bad creating higher resistance.