I may try grounding elsewhere than at the amp.
No, don't do that. The return current from the speaker will want to find its way back to the circuit ground of the amp, and the consequences of that will not be good.
If you connect the negative speaker leads to the ground of a different component in the system, the return path that will be taken to the amp will be through interconnect cables. That will result in crosstalk between that large return current and the low level signal return currents the interconnects normally carry. Also, the interconnects aren't designed to carry high current. I'm not sure exactly what the net result would be, but I would expect that at best it would be a sonic mess, and at worst there would be a feedback-induced oscillation resulting in damage to the system and to your ears.
If you connect the negative speaker leads to an AC safety ground point instead, the consequences would depend on the internal grounding configuration of the amp and the component it is connected to, but I would not expect that to be healthy either.
Regarding the reasons for the sonic benefits you have noted from the cabling arrangement you are presently using, what you are doing is increasing the inductance of the runs. Putting the same signal through two conductors that are physically joined and in close proximity will result in a higher inductance than having those conductors conducting the same current but in opposite directions (i.e., one conducting signal, the other signal return). Also, a run of Romex inherently has much higher inductance than a run of Nordost of the same length. Also, although it is not particularly relevant to your present configuration, increasing the spacing between conductors carrying signal and return currents will result in increased inductance.
The net effect of increased inductance will be to attenuate the extreme upper treble to a slight degree. Exactly how much will depend on the impedance vs. frequency characteristics of the particular speakers, particularly in the upper treble region.
The results were positive,and the increase in detail, less hash, and more rounded full body images has made a believer out of me.
A slight rolloff in the extreme upper treble region, resulting from increased cable inductance, would seem consistent with a perception of greater body and less hash. Perhaps the increase in perceived detail goes hand in hand with there being less hash.
If you haven't already tried it, btw, an interesting experiment would be to interchange what you are doing with the positive and negative runs, i.e., use the Romex runs for the positive connections, and the Nordhost for the negative connections. I wouldn't be surprised if there isn't much difference from the results you are getting now. I would imagine that a difference would be most likely to occur if a solid state amplifier having both wide bandwidth and a feedback loop were being used, due to differences in the amount of rfi pickup in the cables that might enter the feedback loop.