After talking to T. Davinson this afternoon a few things became clearer.

Because of the huge increase in leakage currents, and the 20 MOhm bias resistors uses across most of the pre-amp cards, it is likely that the induced voltage drop has been great enough to not fully deplete 
the detectors.

Maeve and I carries out extended bias curves for most of the detectors and found the plateaus to be reached at much higher voltages than expected.
See detector health spreadsheet and figures in attachment #1.

Bias voltages have now been increased on most detectors to put them in the plateau region.
Some also seem to be shorting to ground.

SiPad #1 on the inside detector was found to like have a short so is now unbiased and its n-side has been removed from the trigger mask.
The trigger mask for FEBEX#3 on x86l-124 is now 0xcf5a (n-side 2, 4, 5, 7, DSSD, crystal PD).

In both detectors:
    The FPGA energy filter SUM_A, GAP, and SUM_B were changed to 100, 200, 101, respectively to better account for the long rise times of our detectors.

    Sparsifying was also turned off for all channels. (How does this interact with data filter control 0x82?)