Dave,
on all of these multi-dial calibrators, you better use full scale output, instead of 10% f.s. only.
Only then, you get full accuracy and stability out of it.
The 10V range (f.s.) is the most stable also.
You can easily see that in the datasheet:
accuracy, 10V: 0.05% of setting + 0.005% of range
stability, 8hrs.: 0.0075% of setting + 0.001% of range
That ..(x% of range) doubles your in-accuracy and instability, if you set these instruments to 10% only.
Therefore, always dial to "10" on the first digit: .(10) 0 0 0 0 0 V or to .9 9 9 9 9 (10) instead of 1. 0 0 0 0 0 V
If you repeat that A.C. stability (noise) measurement on your 34461A with 10V f.s., you would see a much better performance, about ten times better short term stability, I bet.
These AC standards usually convert a programmable D.C. reference to an A.C. output by comparing the D.C. setting to the precision rectified A.C. signal..
Therefore, it's a pity, that the DC to AC comparator / rectifier is not explained in the manual, as the Theory of Operation chapter is missing.
Maybe, that's done by the XR-2228, although its squaring accuracy is normally too bad for that.
That means, that this instrument also contains a D.C. reference standard in first instance.
You may measure the DC reference of 0.. 10.00000V, or so, somewhere inside the circuit.
I bet this D.C. reference part of the circuitry is very similar to the DC standard you also own from KH /EDC.
Concerning the muddy corner of the instrument, that may be caused by high voltages present there.
If you open such HV calibrators, DC or AC, you will always find an accumulation of dust, something like carbon black, at and around all parts, which carry the HV, due to electrostatic attraction.
This one looks a little bit odd, anyhow.
Frank