Sounds nuts to me. The error you calculated assumes the course has no concave sections. In real life the errors would be larger. It's not just the outer corners of the course. Inner curves, or concave sections, would also shorten the distance. On a wiggly road, and even in Florida we have many, the SPR would be much shorter.
What would be the maximum possible difference?
After thinking about it, taking it to extremes, if the SPR was a straight line, and the road wiggled like a snake, taken to the max without disturbing the straight line SPR, the road edge would look like a continuous set of alternating radius curves, with the curve’s radius bing the road with.
Did a little thinking, if my math and theory are correct, in this case the max difference in measured length is independent of road width. It is 2/pi. An error of 1867 meters on a 5K course. That's quite a bit.
Then I thought of another extreme example, that takes the error to infinity. It's a bit silly but what if the course round and round in a circle. How about a tiny circle? Round and round the circle at the end of my road?
The SPF would be just wrapped round the nail in the middle of the circle and the center line distance would be a circle around that at 1/4 the radius of the circle. The math error gets impossible. Also the same course, moved to a different venue with a larger circle, or dumped in the middle of a large parking lot, changes length.
Why are they so silly as to assume that the blader will not take the SPR? I would if I was in front.
It also seems to me that elevation gains and drops would be a lot more critical to records in that sport. How to measure them accurately is an ongoing problem.
I think they need to take the Triathlon approach. Not comparing records on different courses but records on the same course, year to year. The way to settle the record holder is to have a world meet once a year that is run on the same course each time. This also equalizes for the ongoing upgrading of technology, wheels, bearings, titanium and carbon fiber thingies.