130419-2354 EDT
SG-1:
If you insert a low pass filter, like what I described, you should see the polarity change when you switch the meter leads. I don't know why you don't presently, but it probably relates to saturating the meter input with the very large AC waveform compared to the DC magnitude. The time constant of 1 megohm and 0.47 ufd is 0.47 seconds.
You have no need to work with the meter in a low impedance mode.
Somewhere there is phase controlled equipment where the plus and minus turn-on phase angles are not quite the same and thus a DC current is created, or half waving is occurring.
The source of the DC generated current is going to flow thru all the loads and the supply transformer that are connected to that transformer. How much flows where is dependent upon wiring and all the various load impedances.
If you did not have the changing DC voltage, then I might think the source of the current had an open diode or SCR. On the other hand possibly the variation results from various loads being turned on and off.
How to find the source?
This will be isolated to the loads on the secondary of the supply transformer. Suppose this is a 50 kVA 120-0-120 split phase source. Any electronic controlled load with input diodes or SCRs, that is a fairly large power load, is a likely source. Maybe there are many sources. I picked on 50 kVA as an example because that is what supplies my home. But the voltage I mentioned in my previous post would be due to wiring impedance. Since the DC voltage can only occur on the secondary side of a transformer you don't have to look beyond the loads on that secondary.
How can we sense a small DC current superimposed on a large AC current? No good answer. I would try to use a Hall probe, with DC capability, where the maximum AC current does not saturate the Hall device. Use a low pass filter and look for a change in the DC component. My Fluke Y8100 probe on the 200 A range will resolve 100/10000 A (10 mA) on my Fluke 27.
I did not quite beat your 12:30 .
.