Summary:

After engaging resonant gains in MICH, SRCL, and PRCL, DRMI seems keeping alignment with all ADSs (IMMT2, PRM, BS, SRM).

Detail:

To investigate that we need resonant gains in MICH loop quantitatively, I measured the spectrum of MICH, PRCL, and SRCL error signals with ADS EXCs were on while turning off ADS feedbacks (fig1).
Basic idea is that the peaks appeared in MICH error signals multiplied by the coupling function from MICH to POP90/AS34 should be lower than noise floor of POP90/AS34.

According to the coupling function measured in the original post (MICH to POP90: ~0.008 and MICH to AS34:~0.03), we can obtained the following table.

According from the above tble, it is better to put resonant gains to reduce the offset of alignment signals at least for BS yaw, SRM pitch, SRM yaw, IMMT2 pitch, and IMMT2 yaw.
So, I tested the resonant gains.

Unfortunately, MICH UGF is close to 30 Hz, it is difficult to engage resonant gains at 30.125 Hz and 32.125 Hz.
On the other hands, resonant gains can engaged for the oher ADS frequencies
Figure 2 shows the filter bank of MICH1, and FM4 is the resonant gains I implemented.

For the other LSC loops, resonant gains are also implemented as shown in fig3 and 4 because the UGFs for these loops are higher than that of MICH, and we can engage the resonant gains without any problems.

After that, I tested the ADSs with DRMI 1F lock and DRMI seems stable with all ADSs are ON (fig5).
Also, when engaging the ADSs, POP90 and AS34 signals were increased, so ADSs seem to work well.

Note:

Since the resonnt gains at IMMT2 ADS frequencies cannot be engaged, IMMT2 ADS should have an offset.
So, the current lignment is still not the best, so it is better to put resonant gains at MICH, PRCL, and SRCL loops by increasing MICH loop UGF if possible.