That is something one forgets of course, and I certainly tend to forget such things. It is interesting that you would mention 20kHz, because guess what frequency the Multi’s PWM (which it uses to shape the sine wave) runs at … 
It might explain why the ET-Blue garage door opener at my previous place had such a propensity for tripping the RCD, unplugging that seemed to significantly reduce the problem… and the only thing that device had in the back was your usual triangle of MOVs. It never made sense to me… still doesn’t really.
In any case, I found that the clamp meter is amazingly effective and accurate (but it is TRMS!) in debugging these issues. The particular one I have has a peak-hold function that scans fast enough to see exactly how high the leakage rose. What is really nice about that is you can truly divide and conquer, disconnect circuits and observe the effect. It adds a lot of speed to the process.
I even did one test where I removed the RCD and replaced it with an isolator, then let the inverter switch over, and I recorded a 40mA peak! And all of that was transient behaviour from the collective fleet of appliances… because the rest of the installation checked out fine. Earth was proper, earth continuity was good, standing loss was acceptable…
I ended up spending around 9k on solving that problem. Half on the clamp meter, half on the RCD. I was really really fed-up by that point, and practically threw money at it!