Having decided to go with the 0.3pF + 976k correction circuit for this first batch, we will be characterizing it to see how consistent it is and look into hardware and DSP options to get the flatness to be more in line with other scopes. Below is a screenshot of the issue with a 100kHz square wave:
Below is the ngscopeclient scopesession file with associated data for this capture:
9Vpp_100kHz_square_10V_range.zip
Here's a amplitude vs frequency plot taken end-to-end (note the artifact at 1MHz is due to the function gen, not the front end):
config 18 - 20 Vpp - attenuator True.csv