One thing I tried that was a bit more revealing w/regard to the actual ripple coming out of my PSU was to trigger the scope using one of the switching signals on the inverter board. I still had a bunch of other noise coming in from elsewhere so I put the scope into its averaging mode. Then I started seeing ripple I could start to believe in. It's about 7mV peak to peak. Thing is, the time frame is about 2us for that much delta-V. Going into how much current has to flow into the filter capacitors to produce that voltage change in 2 microseconds yields an absurd amount of current. The only current draw is the 100 megohm voltage divider used in my feedback loop. So I think I'm seeing the circuit ground(s) bouncing around, probably due to switching currents in the inverter.
I think you have it right!
Keep in mind that some SMPSUs actually need some load current beyond what a 100 megohm load provides, just to get their control loop into operation. The energy storage capacitor / low pass way is not effective against higher frequency bumps, and by basic nature, if it has to supply currents, it will inevitably make ripple. As you say, getting at the switcher ripple by using capacitors would involve high currents, even if for very short times.
The more absolute way is to use a post regulator linear regulator that has a very high bandwidth, so fast that it can follow the ripple transients, and regulate them away. The problem then becomes making a very clean reference.
I have also been disappointed by some commercial mains line filters, until the day when I was providing generator support to some radio enthusiasts field day, when I was shown the one that also choked the ground line. Those guys could make the contribution from power line interference become hard to detect, even on a field with full blast HF transmitters coupling directly into them.
Extra regulator stages may not be conveniently practical, and is a different thing to deal with than power line conducted rubbish from LED supplies designed down to a price. The rules for these might be different in USA than we have in Europe. Regardless, once you know where the currents are, and how they interfere, then you can definitely stop them.
It gets desperate if we have to reach to more complicated and expensive low ripple supplies, like a highly developed Ćuk converter. Also, the effective way to stop circuit grounds "bouncing things around" is to deny these routes the currents. There is always some inductive impedance in the ground connection, and the idea that "ground" is some kind of massive sink that can stay at 0V no matter what we try to dump into it is just wrong. For me, the switcher currents usually involved motors, in one project, a pair at more than 70kW each. Stopping noise had to be the most frustrating aspect among all of it!
