Before May 2000, the U.S. government used to degrade the civilian GPS signals, resulting in horizontal errors of about 100 m. At midnight on 1 May 2000, this process, known as selective availability (SA), was permanently turned off (read about it at the National Geodetic Survey site).
One way to deal with SA was to record the GPS coordinates of a location over a period of time and then average them. While going over my old field books, I found the GPS coordinates of some of my collection sites recorded before and after SA was turned off.
At station AX-2, using my Garmin eTrex receiver, I recorded 5 pairs of latitude and longitude over a 25-minute period on 23 April 2000 and an additional 8 pairs over 40 minutes on 30 April 2000. If I had known that SA was going to be turned off in a few days, I wouldn't have wasted my time, of course. Here are those 13 pairs of coordinates, A thru L, marked on a Google Earth picture of the site. The 13th marker, which would be M, overlaps that of the yellow marker with the star.
The collection site was between a narrow paved road and a pair of railroad tracks. The vertical white line is 100 m long. The distance between the westernmost and the easternmost spots, C and L, is about 54 m, while the distance between the northernmost and the southernmost spots, K and J, is about 95 m.
On 2 May 2000, I went back to the same site and took 4 more pairs of readings. The signal was so stable that the 4 pairs had the same latitude and differed by only 0.1 seconds in the longitude readings, resulting in only 2 pairs of different coordinates. Those are indicated with the yellow and green markers with stars. The difference between them is about 2.5 m. I was probably standing on the road right there somewhere in the vicinity of those markers while taking the measurements.
The average coordinates from before SA was turned off were close but not identical to those after SA was turned off. Interestingly, the very last set of coordinates from 30 May 2000 (marker M) were identical to one set from 2 May. Also, I noticed while looking at the above picture that if a line is drawn between K and J and another one between C and L, their intersection lies close enough to the actual location of the station. There is probably some probabilistic explanation behind that.