I did a check of the darks on a single detector (6, 1_44) by pushing 1 dark from each month that had dark exposures available through the cpVerifyDark code. This tests the 3-input dark as well as ~15 others. None of these would be accepted for use today. The NOISE test seems to be consistently failing (measured noise after overscan/bias/dark consistent with read noise), the CR_NOISE test fails for about half (same test, but with CR rejection applied), and there are a handful that fail on PROCESSING (CR rejection could not run due to too many detected CRs). Only one of these darks has 8 inputs, with the majority having only 5 inputs.
I had hoped to directly check the inputs, but those inputs are not available at the SDF for the 3-input dark, leading to my scan approach.
I've uploaded the cp_verify notebook I used to check the darks, the commands used to run that verify process, and a quick table of the darks studied, and the number of inputs used in each of those darks.