Using iftarget_tests, configured Q-band to observe 2 lines each in their own VEGAS bank, 40.6 and 43.6 GHz (~3GHz spread)
This put the IF Rack attenuation at 18dB (giving an output power of ~1V (as expected). The converter rack then was balanced from 25-4dB.
When only observing at 1 frequency (42.6 GHz), the IF balanced at 19dB (1V), and the converter rack at 10 and 18 dB (2 and 1V for left and right circular respectively).
When splitting the frequencies in test 1 to 3.5GHz, one attenuator dropped to 2.875dB, suggesting we are approaching the limits of these capabilities, but not there yet. The 6GHz bandwidth possibility may be suspect.