GPS repeat period calculated from daily GPS almanac.
Nominal value is 86164 seconds (sidereal day) not 86400 seconds (solar day). For 32 SV times 365 days times 12 years the mean value is 86155 seconds. Kristine Larson has papers on this. But that's not the interesting part.
Instead of bulk averaging, let's see how the repeat time changes over the span of months and years. In the plot below, the period is plotted over time (up to 12 years). The x-axis is one point per day, from 2001 to 2012. The y-axis is centered at 86150 seconds with 10 seconds per division.
PRN04
I would guess the ramps are slow orbit decay and the occasional jumps are station keeping burns, about once a year. But even that's not the interesting part.
Below are 32 plots, one per PRN, covering up to 12 years each. Not all SV have complete data sets so some plots are less than full width. The y-axis is manually clipped in some cases (for points far above or below the plot). It's amazing how different each of the SV are. Sudden changes in character might be SVN/PRN reassignments, e.g. PRN06. These are quick automated plots; I can get fancy later with an aligned 32x365x12 matrix.
PRN01
PRN02
PRN03
PRN04
PRN05
PRN06
PRN07
PRN08
PRN09
PRN10
PRN11
PRN12
PRN13
PRN14
PRN15
PRN16
PRN17
PRN18
PRN19
PRN20
PRN21
PRN22
PRN23
PRN24
PRN25
PRN26
PRN27
PRN28
PRN29
PRN30
PRN31
PRN32