I was intrigued by the stamps shown. I dug out my used copies of 1702 and 1703. They had been sorted previously by myself and many other collectors before me, likely only by the visual differences of snowflaking, etc. I confirmed the ID with a UV light and corrected a few mis-IDs. I found:
1702 with overall tagging - 117 copies
1703 with block tagging - 110 copies
(To my surprise, I think I found one without tagging, 1703c)
I did find 6 of the 117 copies of 1702 with a slight upward and lighter partial doubled lettering, roughly 5%. Three of the six very very minor. The most prominent being a pair, the right stamp shown here:
I looked through the "U.S. Specialist", but did not see any articles on the printing of this specific issue. Before going on, I would note an earlier post which said, in part:
double impression or re-entry ( whichever you'd prefer )
There is a difference in the terminology as these two terms are not interchangeable. Although the result may look alike! Re-entry is a doubling on the plate which would create many identical impressions over a press run, and yes, typically proven by a confirming copy. A double impression is a unique occurrence of the plate/paper interface where each similar occurrence can/will be slightly different.
Going on ... It is clear to me this minor doubling is not a plate defect, but rather a printing effect. Consider, after all, this is a rotary process going how many press cylinder rotations per minute and how many hundreds feet of paper pass through the press every minute? (I would love to know the rates more precisely.) It doesn't take much chatter or vibration of the paper web, or slight mis-synchronization of all the moving parts at these speeds to have a tiny difference or defect show up in a small percentage of the final product. And across the web of paper affecting multiple positions which see find as pairs, blocks, etc.
On a tangent, I found other copies of 1702 with slight transient color mis-registrations, maybe also about 5%, not enough to be out of production tolerance nor catalog listable as constant.