Figure 1. Shows the test of changing the modulation depth of sidebands. In purple with 12dBm on the 6MHz, and in blue/red with 15dBm. What is surprising is that the spectrum of B2 8MHz increase with a higher modulation depth. With a noise in a shape of a pole at ~500Hz increasing.

A possible explanation is that the modulation of the 6MHz adds phase noise at the 8MHz frequency. Are the two sidebands modulated using the same crystal?

This could be also an explanation for the SSFS error signal and quadrature being affected by a noise shape like ~500Hz pole. If the reverse process happens, the 8MHz modulation creates a phase noise at 6MHz. A simple check of that hypothesis is to increase or decrease the 8MHz modulation depth by ~3dB and see if that changes the spectrum of the SSFS quadrature signal. There might be already past tests with such a change that just need analyzing.

Note that we don't have a good way of measuring the 6MHz phase directly, as all the photodiodes on B2 and B4 have a notch at 12MHz.