Charles, if I may add my 2-cents ...
Looking at your 8 hour (scaled to 24 hrs) results:
Witschi: 1.6 +/- 0.5 s/d
WatchTracker: 1.8 +/ - x.x s/d
x.x (error) not given
The results are 100 % identical.
For data taking and analysis it is MANDATORY to think about syst. and stat. measurement errors.
Btw, most complicated to measure is 0 (Zero).
The 'dips' in your 3235 amplitude and rate curves I have also seen before for my 3235 SD43:
They explain (in my opinion) why your 72 hour long PR measurements where not smooth:
Amplitude values seem to depend (a bit) on the exact time you measure during the PR.
I observed a similar effect for my 126600 SD43:
PR measurements for 3130 and 3187 seem to be smoother (no significant amplitude dips) than for your 3235:
But there is very little systematic work done on this 32xx/31xx comparison.
An interesting question might be how reproducible these dips are. It seems that they have no impact on excellent timekeeping.
Hence it's more an academic exercise.