Wow, interesting device. I always assumed it would be something like Fluke 5450A (i.e. a bunch of relays and resistors), but here on the photo I've noticed a LM399, which seemed odd.
So I started reading the theory of operation from the manual and it is very interesting. It is just 6 reference resistors, the rest are analog control loops to draw current and provide the correct voltage drop - I guess not dissimilar to a DC load in the CR mode (chepos implement that in the microcontroller, some Agilent loads use analog multipliers).
I would never think that with such technique the specs can be so good - for 10K: 90 days 7ppm + 50 uOhm, 1.5 ppm tempco. Actually, not much worse than 5450A.
Thanks for the photos.