From original post:

"As it turns out, the voltage is coming from the ground (bare) wire from the #6-3 feeder from the service panel - even when the house is completely disconnected from the power line. BTW - the sub-panel has two buses - the ground wires are connected to one and the neutral wires to the other - When we connected the #6-3 ground to the neutral bus the shock stopped - but that doesn't meet code."

The crux of the problem is in this statement. Both neutral and ground should be tied together at the main panel, and left separate at sub-panels. I'm not an electrician, but does #6-3 imply three wires #6, or 3 wires #6 and a ground? If it the last one, is the gauge of the ground wire the same (#6)? If we can eliminate the problem by jumping neutral and ground at the sub panel (not code), we should look upstream to the main panel. Are the two connected properly there? If so, take a meter from that point to a nearby ground rod (could be temporary 12" rod for this test). Is there any voltage from rod to main panel junction of neutral an ground? See illustration at the following link:

http://edis.ifas.ufl.edu/ae019

The dairy industry has similar issues where cows are extremely sensitive to stray electrical currents flowing in the soil, or concrete.

You might need a substantial buried ground plate(s) at the main panel. See link here:

http://www.erico.com/products/GroundPlates.asp

Jim