Interesting to measure the emissions of a device that works by measuring the emissions of other devices. Something like using the a spectrum analyzer to measure a spectrum analyzer. LOL I am glad to see this. Have you looked up the ESR of the decoupling capacitors? This is some times a place to start looking? I haven't had a chance to look at the schematic and layout. If there is some long or exposed clock lines could be moved into inner layers or adding some very very small capacitance per length. Another thing to look at is matched impedance of all the clock and data traces to the input and output impedance of the lines.
thewyliestcoyote Thanks for the hints! There was indeed a clock line on the bottom layer... I was able to get emissions down by adding the series termination (luckily had added the footprint for that, but when built the board just mounted a 1 ohm resistor). I also added ground flood on the bottom layer... the prototype didn't have that (relied only on the inner layer plane). I'm trying to change the board as little as possible, as my next order will be for qty 1000! So worried about introducing mistakes...