After confirming that the readout path output of COMM PLL (for CM board) has a large offset (about 4.8V) regardless of the PLL board setting while the output for VCO path was seemingly OK, we pulled out the chassis.

It was U37 (AD829 for z1.6Hz:p40Hz) that was busted. 

Jax tested the board, replaced the broken chip and tested again. She also found that the VCO path output was only routed to the external connector on the back panel, so she connected the internal VCO path SMA cable.

The fixed unit was put in place, the offset was not crazy (about 430mV when input was off, 13dB gain, two common filters on, went down to 13mV when both of the common filters were turned off), we cabled it up.