The attenuator switch as a problem is now less likely. See S1A(F) (page 6.3/6.4). One can see the S1A(F) contacts resistor string S1R13 thru S1R18. As S1A(F) proceeds from 3uv to 1v, the range switch removes one resistor at a time ending at S1R17. At this point S1R18 resistor is the only resistor left in the DC feedback loop. Also, up to this point there has always been resistance in the path for DC feedback. Starting with 3mv range, the R18 resistor is removed, so the feedback loop is now a short, and instrument oscillates. Move the attenuator switch to 10mv. Now there is NO feedback on the SIA(F) portion of the switch, and no oscillation. All 3mv,30mv...30v,300v positions are shorted to feedback, and all 1mv,10mv...1000v are no feedback. Thus, with feedback closed loop shorted from S1A(F), the instrument oscillates. No feedback, no oscillation. I still don't know why this is happening, but now I see the condition that causes it.
There is a possible hint in the manual, Theory of Operation, page 4-2, paragraph 4-24, about the gain of the AC amplifier.
"AC feedback from the emitter of Q3 to the emitter of Q1 is used to vary the the gain of Q1 through Q3...in the ranges 3µ - 1mV, resistor R7 is shorted out, decreasing the negative feedback applied to Q1."
That means, in the upper ranges from 3mV onwards, the AC gain should be reduced as less total open loop gain is required.
If that doesn't work correctly, the overall amplifier may oscillate.
Therefore, please check S1B(R), if that shorts R6 correctly to GND in position 1 - 6 only.
Is C2 i.O., does it still have 200µF and low ESR value? Same goes for C6, C7, which might also influence AC gain.
I assume, that you already checked all other electrolytics, which smooth the supply voltages.
Are the expected DC level around Q1..Q5 ok, and all resistor values, R1..R22?
What's still strange is that this oscillation starts after warm-up, so it may have something to do with the neons also, maybe you try anyhow what happens if you desolder C1 on A3 board.
Frank