A more detailed alog with analysis will follow.

I injected some 10-100 Hz noise in CHARD_Y. It looks like we have a safety factor of 30-100 for noise coupling above 10 Hz.

Also from previous OLG measurements and models, it looked like we could simply increase CHARD_Y gain by a factor 3 and improve the residual motion.
It worked. I tested a gain 3x the nominal for ten minutes.

CHARD_Y noise ampl 20  butter("BandPass",6,10,100)

start   PDT: 2023-07-26 14:20:47.869550 PDT
        UTC: 2023-07-26 21:20:47.869550 UTC
        GPS: 1374441665.869550
stop    PDT: 2023-07-26 14:23:11.548956 PDT
        UTC: 2023-07-26 21:23:11.548956 UTC
        GPS: 1374441809.548956

CHARD_Y gain 180
from    PDT: 2023-07-26 14:28:47.965492 PDT
        UTC: 2023-07-26 21:28:47.965492 UTC
        GPS: 1374442145.965492
to      PDT: 2023-07-26 14:39:08.938636 PDT
        UTC: 2023-07-26 21:39:08.938636 UTC
        GPS: 1374442766.938636

CHARD_Y gain back to nominal 60
from    PDT: 2023-07-26 14:40:25.797705 PDT
        UTC: 2023-07-26 21:40:25.797705 UTC
        GPS: 1374442843.797705
to      PDT: 2023-07-26 14:50:34.172825 PDT
        UTC: 2023-07-26 21:50:34.172825 UTC
        GPS: 1374443452.172825