Your zener reference could take a few weeks now to get stable. If you calibrate it now, you should compare it against your 10V reference every day until the readings remain fairly constant.
It took several weeks before my meter would stay close to zero with the inputs shorted with a copper wire across the inputs. I ran ACAL ALL a couple of times a day for several weeks, and I could see the zero drift less with each passing week. Temperature is also a big factor too. This thing is very sensitive to temp changes, even by a change of 1 deg. C
That's absolutely not right, or not in that extreme, sorry.
The LTZ reference itself is stable within a few minutes. Afterwards it may drift a few tenths of ppm, or less. That's how my two external LTZ1000 behave @ 45°C.
For the 3458A it's hard to tell, how the LTZ behaves at 95°C, because normally you cannot measure the internal ref from outside, and the intial drift is determined totally by the gain resistors drift.
The 3458A as a whole only needs those 2..4h hours of warming up, 12..15°C max. difference of internal vs. room temperature.
Take care, that the difference is not greater, otherwise clean the fan filter. And put it on desk top, not in a rack for best metrological performance.
I always shut off the 3458A after a session.
When I power on again, it takes about 2h, and perhaps (not necessarily) an AUTOCAL, and then the instrument is back to a few 10nV zero reading, and Full Scale reading to tenths of ppm.
That does not take weeks, that's absolutely not correct!
Even after weeks of being shut off.
For precision measurements, half a day of stabilization is used.
Afterwards, Autocal is only necessary if Rt changes more than 1°C.
But you can measure such small effects only, if you have absolutely stable RT over a longer period of time, and if you have a source with equal or less drift than the internal one.
A Fluke 721A is not sufficient.
LM399H is minimum requirement, better another LTZ1000 or oven stabilized SZA263 (Fluke 732, 5440, 5720, 7000 etc.) reference is required.
Also a high quality short is required, either good laboratory cables, or a clean copper wire.
If the instrument is not rock stable under the described conditions, it would be faulty.
How long the Geller ref takes to stabilize, I cannot tell. But weeks would be too long, to give a reliable reading, I think.
Frank