Did you by any chance throw an amp clamp around the entire vibrating conduit? Were your harmonic measurements voltage, current, or both?
On harmonics: remember that non-linear loads consume harmonic _currents_. Like any load current, harmonic load currents cause voltage drop in the supply impedance. In the same way that a resistive load will drop the fundamental supply voltage, non-linear loads will drop the 'harmonic supply voltage'. This change in supply voltage shows up as harmonic distortion of the supply voltage.
With a perfectly 'stiff' supply, non-linear loads would cause no distortion in the supply voltage. Harmonic currents would flow, but the supply would provide them with no drop, so the supply voltage and waveform shape would remain constant. Conversely, a _stiff_ supply will mean greater harmonic current flow to the non-linear loads, because it has less voltage drop.
Hopefully someone here has seen this sort of thing before. I don't work with large UPS systems, and don't know how their switching output might couple to bonded metal (with the VFD motor systems that I do work with, the high frequency switching capacitively couples to ground, so even without an actual ground fault, you do see current flow on bonded metal, and you will measure 'ground fault' current).
-Jon
