How linear are the SiPMs?

Three things that can impact SiPM linearity are saturation, crosstalk, and afterpulsing. Saturation is related to the total number of microcells sharing a single biased cell and the recovery time of the cell, crosstalk is the emission of photons that trigger other cells in the SiPM, and afterpulsing is an additional pulse triggered by charge carriers which become trapped within a cell (otherwise known as a correlated delayed pulse).

See Table 1 on page 1 for the ranges of photons our SiPM detectors can handle without saturating:
https://www.onsemi.com/pub/Collateral/AND9776-D.PDF

I understand this to refer to the full sensor (i.e. not the microcell). Since our array has 64 sensors on it, light that is distributed perfectly uniformly over the surface will deviate 10% from linearity for fluxes ranging from 4000 to 1900000 photons, depending on the sensor. Fortunately, it looks like the 6mm sensors on the array we just recently installed (ARRAYC-60035064P-PCB) will mean that 1900000 photons are needed to cause a 10% deviation from linearity assuming perfectly even light distribution.

I calculate that about 14000 photons are expected from a 10 MeV electron in the scintillator we are using (BC-404), so that would imply that the rate of photons shouldn't affect linearity very much, but John informs me that 1% could be significant. It is not yet clear whether crosstalk or afterpulsing could potentially cause a problem.