I identified an issue with the Pcal boards, and needed to find out what the impact is.
I modeled the boards with a photo diode in SPICE to see look at the frequency response of the output voltage. The issue with having the wrong capacitor manifests itself as rather intense gain peaking at around 1 MHz, and various instruments that Pcal oversees have different versions of the board which are all impacted differently.
Of concern to LIGO is how bad is this <5 kHz?
Well, with the current capacitor (in this 2_7_pf.png attachment) you see the response 5 kHz is roughly 8 * 10^-4 % higher.
With a replaced capacitor (in this 5_0_pf.png attachment) you see the response 5 kHz is roughly 4 * 10^-4 % higher.
Moral of the story: since Pcal isn't used above this kind of frequency, there is no need to worry about this issue since the error is miniscule in these frequencies.
Overall, regardless of fixing the capacitors, the response will be accurate to within 1 % up to above 100 kHz.
The other aspect that may give some frequency dependant error is coherence time in the sphere:
There is an equation listed in integegrating sphere technical notes, as well as presentations by NASA [page 4] that looks at exponenetial time decay of pulsed signals, this can be used to get an apporximation for the coherence time constant within the integrating spheres. From that, the bandwidth (FWHM linewidth) of the sphere is: 1/ (pi * tor)
It can be calculated that the sphere itself is good within 1 % up to 2.14 MHz.
So the electronics boards are the main limiting factor.