The reason why the IFO second loop didn't like the ISS 2nd loop is because the ISS 2nd loop offset is now hard coded (28076) while the power and diffraction both changed because of the diode swap and subsequent tune-up.
I remeasured it at 40W (H1:PSL-ISS_SECONDLOOP_REF_SIGNAL_ANA=-0.027689945 for 16-17% diffraction) but Vern told me that it was decided that we'll run IFO at 25W during ER9, so I measured it yet again at 25W (-0.027651985).
I'm puzzled that the number changed this much (previously it was -0.9826934814453125 at 40W).
It's also odd that there's not much difference between 40W and 25W. The diodes are definitely connected, the readback of analog sum (H1:PSL-ISS_SECONDLOOP_PD_14_SUM_OUT) is almost 100% coherent with the digital sum of individual PDs (H1:PSL-ISS_SECONDLOOP_SUM14_AC_OUT and H1:PSL-ISS_SECONDLOOP_SUM58_AC_OUTPUT).
Anyway IMC_LOCK guardian was modified for 25W (H1:PSL-ISS_SECONDLOOP_REF_SIGNAL_ANA, iss_diffracted_power_target=16.5).
Update: I was likely tricked by MEDM screen (graphics of switch states sometimes don't agree with reality) when I was doing the above measurement with only MC locked.
With full IFO the above offset was found totally off, and at 25W H1:PSL-ISS_SECONDLOOP_REF_SIGNAL_ANA=-0.5885 or so for 15%-ish diffraction. Didn't have time to remeasure at 40W.
Summary:
The 2nd loop engagement logic is bad as it wastes too much time waiting for a luck, but waiting for a luck doesn't do anything good.
Details:
2nd loop offset servo can take the 2nd loop board output or the diffracted power as the error signal.
At 40W or 25W, without engaging the 2nd loop, the output of the 2nd loop board always goes rail to rail even if the offset is correctly set just because the error signal is big.
Despite this, the offset adjustment servo is engaged anyway using the 2nd loop board output. The board bang-bangs forever, but eventually the guardian grabs a lucky moment when the board output happens to be small enough of a number, and thinks (incorrectly) that the offset servo converges. And then it engages the second loop. But this is as good as nothing IF you know that your static offset is reasonable.
Until a better criteria to engage the 2nd loop is found, I think the best strategy is to
I changed the guardian sans step 3. above:
It works.
2nd loop sudden death problem:
Jenne found that ISS 2nd loop is suddenly disengaged because the 2nd loop board output exceeds the OFF trigger threshold of 5 (first attachment), killing IFO.
The second board output goes close to 5 kind of often now, it seems. Since we don't have time to do a good investigation, for the moment I set the threshold to 10 (which sounds too large) and see how it goes.
In the last lock the IFO survived with ISS 2nd loop on for 10 minutes. The lock loss didn't seem to be due to ISS (2nd attachment).