Got to a 40 minute lock tonight. All other locks broke because of me poking something.
I redid some sensing excitations, right after carefully measuring the CARM OLG at its excitation frequency, so I can get at the open loop PD response.
I also used a MCL feedforward filter of Ignacio's which did not inject any observable noise into the CARM error signal during PRFPMI lock. He will make some elog about this.