I measured the open loop transfer function of MICH, PRCL and SRCL when DRMI is locked on POP (full resonance). It turned out that MICH and SRCL transfer functions showed some weird features at high frequency, due to cross-coupling to PRCL.

I retuned the POP9_I > SRCL sensing matrix element from -0.0467 to -0.02, and added a matrix element from POP9_I to MICH with a value of 0.009. In this way both MICH and SRCL open loop transfer functions have a smooth shape as expected.

For reference the table below shows the UGF and measured instability frequencies of all three loops (if measured)