8904 HP Multifunction Synth Floating impedance disparity

[coppercone2]

https://www.artisantg.com/info/HP_8904A_Service_Manual.pdf

Page 135 has the circuit

I have two of these units, when floated one unit measures some megaohm to ground, the other unit measures 1.000 K to ground. It looks like its measuring the R423 resistor behind the CM choke, while the other one is not, at least thats my hunch. Both boards look the same on thermal and it looks like the correct relays are getting warm on float switching.

I am probobly missing something simple, but why is there such a big difference in input impedance? Both units have 2 output cards, so 3/4 cards measure 1 megaohm, while 1/4 cards measure 1k.

If I set it on a 5Vpp output and measure sine wave while its floating, 3/4 channels measure 1 megaohm and the other channel reads 300 ohms to earthground (dancing around), but the output voltage is correct

where should I look?

so in short
1) channel 2 on unit #2 has a 100x lower isolation impedance when measured with DC ohmmeter
2) isolation impedance on channel 2 unit #2 has a isolation voltage that seems dependent on output voltage setting of output amplifier

[coppercone2]

or a real test on float impedance would be nice

i thought to measure voltage drop on a ground ref resistor?

[coppercone2]

when I put a 10 ohm resistor across the output the float button does the same thing and same attenuation ratio to both channels. I am guessing there is something fundamentally wrong with using a DC DMM to measure resistance on the output stage for some reason even if the result seems convincing, because I tried it at DC and 50Khz sine wave and the isolation seems very similar (impressively so).

Don't know why 1 instrument does this, maybe something with the filter or attenuation, they have a slightly different PCBs where the 2nd output PCB has a jumper on it but the other one does not for some reason.. the jumper is involved with the isolation when measuring it

I was not expecting 'good' results from a DMM but I thought exactly 1Kohm vs 1.1Megohm was a little suspicious.. maybe its soft overloading some stage that lets it measure true DC impedance on one channel because of happenstance of component values and setpoints? whereas on the other channels its somehow fighting the amplifiers.

Any ideas on if there is a better test the voltage drop on output resistor to verify the float?