In real life I don't have a wheel calibrated to 100th of a revolution. I have a road bike, which wobbles when I start up, but gets smooth, stabilized and straight at about 10 miles an hour, well after the first 75 feet.
I use a jones counter that has a bit of whiplash, if I roll forward or back when getting lined up with the start, which is something I tend to do.
I have also noticed that with the same bike and same moisture there is some variation on my cal rides depending on the wind, and more surprisingly the time of day. In the evenings, as the light drops off it takes more counts to get to the other end.
I also notice that it takes about two to four rides of the cal course before the bike and I get warmed up and the counts get consistent.
Since the cal course is 1/2 mile long, thats about 1 to 2 miles of warm up.
While in theory if you multiply anything by 1600 it is mathematically 1600 times larger. In practice, when using an intermediary measuring tool to scale it up you will just scale in and scale up all the errors. That's why blue prints are large and dimensions are printed on them to discourage scaling up from the drawing.
I have no doubt that under ideal conditions an experienced measurer can sale a 25m course up to a Marathon without introducing too much error.
The key is, our method should work for not just the experienced and careful measurer, but for almost any tom dick and harry who follow the process correctly.
Therefore the process should be one that tends to naturally lesson errors, not one that gives an opportunity for compounding them unnecessarily.
I realize that you are not talking about the recommended method for measurement, using the jones counter, but experimental methods using the operators own improvisation of marked wheel rims and electronic revolution counters. While this is technically interesting, it's not providing a bullet proof method that a man in Botswana can replicate, out of a filed guide.
As a person who works in the high-tech field I welcome appropriate use of technology, but often wonder at the lengths people go to modernize or fix something that is not broken. The point of the measurement system we use is to replicate how a rider will ride when on the course. Wobbles, looking up, and down, and all.
The less like the course it gets the less the ride duplicates the variation that the rider will have when riding the real course.
I suppose the logical next step is to eliminate the errors caused by actually 'gliding' the bike over the 25m, instead we will just have to transcendentally think about a meter.
Heck, at only 25 meters it seems to me that you would get equally consistent results with a surveyors rolling push wheel! The point is we know that even when a surveyors push wheel gives consistent results over 25m, you can't reliably scale that up into a marathon.