I don't see what I left out, except possibly the well-known trick of using a voltmeter with known resistance to measure small currents. I put the meter in series with power supply and DUT and let the meter's input resistance be the ammeter shunt. With a 10 Meg meter, 10mV is 1nA. Since I had three DUTs in play, I used three meters, a Fluke 8000A, a Fluke 8100B, and a Monsanto 2000. After measuring the leakage, I shorted the meters. Then I turned off the power supply, which has a few kilohms of bleed at the terminals, and after 10 seconds I pulled off the meter shorts and watched the readings go negative, level off, and fall. This isn't exactly like the IEC standard since the meters load the recovering caps, but since all I wanted was a comparison it doesn't matter. It was quite a surprise to see the ceramic cap perform worse than the electrolytics!
I see that the typical leakage of the polymer caps, like that of the regular caps, is far below the spec sheet limit.