@lars
I love questions, they focus my ignorance. I can describe the reference I am testing - owner is cellularmitosis I believe.
Geller Labs Reference - model number is not obvious in the available view. I don't want to disturb it during a run.
The number 977 is inked on the board. (Or possibly 911 if written Euro-style, but the characters in question look more North American than the Euro style 1.) Don't know if this is the serial number.
The key chip is clearly marked AD587LQ with a date code of 0054. I haven't found a data sheet admitting to the existence of an LQ grade so don't know what that means.
The DUT was exposed to normal diurnal temperature swing (started at approximately peak daytime temperature, and went 14 hours, a little past where temperatures began rising the next day. NPLC for 3457 was 100, 3456 was at default value of 10 (another mistake that has been corrected for the current run.) The DVMs were all thermally isolated from these changes in a room whose temperature was 26 C +/- about a degree. Measurements were taken about every 8 seconds. The plotted data is the running average of 11 of these measurements.
My comment about the instruments being off was a lazy reference to the difference in readings and the difference from the stated perfection of the reference. I have not yet even plotted error bands or the range specified by HP.