Thanks for looking at this.
Basically I had similar thought on this, thermocouple effects, material combinations, DMM internal temp differences... I was not sure about the "in spec" temperature caused deviations, thanks for pointing a this.
Btw. switching the direction of the shorted leads makes no difference.(If its a thermocouple effect of the leads, it should switch the sign in the reading also, right??)
I see that these effects make the µV range kind of unpredictable, (see below) when temperature is changing.
A bit unpleasent that on the "high-end" meter the effect was so obvious noticeable, where the older/cheaper DMMs needs to be treated much harder to show similar effects:
Just done a quick test again:
U1252B: (btw it is an Agilent)
20.5°C::0µV
17.1°C::5µV
U1282A
21.8°C::2µV
17.1°C::37µV
And meanwhile the U1252B is at 18.7°C::-4µV
the U1282A is at 18.3°C::-39µV it has drifted 76µV