Well, I got the boards back with the new buck converter design, and it's a truly epic fail.

The 5v regulator is fine, but it's a tried and true design. The problematic part is the 15V buck regulator. At 1A of draw, it has a good 300 mV p-p of ripple, and that ripple looks like a tempest-tossed sea - almost literally like UUUUUUUU - on the scope.

Thanks to the comments on this log, I was inspired to go back to DigiKey and look for a better buck regulator - preferably one with a built-in switch that could produce up to 2A. The answer looks like it's going to be the TPS54231.

I've also decided to add a U.FL connector on the board and a solder jumper to switch between using it and using pin 7 of the DB9 for the 10 MHz input from the oscillator. This will allow supporting 5680A units that have an SMA output jack rather than having output on the DB9. By special request of one of the prototype testers, I've also added solder jumpers to allow selecting between TTL and RS-232 output on pins 8 and 9, on the off chance that you have an oscillator that has TTL levels on its serial I/O.