I already reported a such kind of error on 5$, repeating most of the time every 3 notes.
The explanation about the three printing drums could be good, but I don't see the point to put the three runs one of each in sequence. and it would need three printing machine...
My explanation relies on the fact that the tiny numbers are Position numbers, not Plate numbers.
So I would think that there is effectively three sets of "plates", but put on the same printing drum. I think that because if you roll up a complete sheet, the diameter of the roll is quite small, (less than one feet), corresponding to a too small drum, probably too small to be usable for the printers.
But if you put three sets on the same drum, you'll need a drum about 3 feet in diameter, much more usable. And think that each "plate" must be curved to be put on the drum. On a small drum, the plate must be quite curved (maybe too much to do a good printing), but on a large drum, the curvature is much less pronounced.
And you just have to sychronize the feeder to follow the rotation of the drum..If something sticks on one of the "plates" only the sheet printed by that plate will show the defect (no ink for instance), the two others will be OK.
And why there is some miss in this disposition (sometimes, you have only one good note between two errors, or three or even four notes), since the serials are printed in the last printing step, nothing tells you that there was no sheet taken out after one of the preceding step by the QC...