Hello,
in general, to determine the drift rate / stability of any artefact standards (e.g. prototype kilogram, Weston cells, prototype meter, Thomson or Reichsanstalt Ohm, OCXO, MASER or Cs clock, etc.) you might always use at least one other reference standard of the same stability class.
From the relative drift of these two artefacts, you can always estimate the drift rate of these references, you might assign about half of the total drift to each individual reference.
The more artefacts you have in your group, the better the estimate of the drift rate.
3 references lets you decide, which of the three is the one which drifts the most.
The drift mechanisms and systematic errors should be well understood, and additional drift influences, like temperature coefficients and so on should also be considered or mitigated.
So it is not necessary to have a superior standard to determine the drift rate.
That procedure is good metrological practice, especially in a certain case, where artefacts are still in use, and no superior standard exists.
That is currently the prototype kilogram, where the metrologists simply compared "Le Grand K' against its about 30 copies, and determined, that the whole group drifts apart, and especially, that the prototype kilogram in Sèvres is the most unstable (rate -5*10^-8 in 100 years)
So that self-monitoring procedure I have used for the last 8 years, on my voltage and resistance references.
I maintain 4 different voltage references, 2 LTZ1000 @ 45°C, continuously powered (systematic drift: typ. -0.8ppm/yr.), the LTZ1000A @ 65°C reference inside my 3458A, not constantly powered (systematic drift: typ. -2ppm/year at continuous power, therefore typ. < 1ppm/yr. due to long non-powered state), and a double SZA263A reference @ 55°C inside my 5442A, also not continuously powered, (typ. + 1..2ppm/yr.)
These typical drift rates are taken from different investigations or datasheets of similar references, like Spreadbury, Pickering, Fluke, Metron, etc.
The 3458A is taken as the baseline; all absolute measurements are done relative to it.
The difference to the 5442A was always less than +1.5ppm over the last 4 years, and the difference relative to both of the LTZ1000 was always less than -3ppm over that time. Therefore, the drift rate inside the whole group is for sure < 1ppm /yr.
A similar procedure is done on 5 VHP202Z, 10.000kOhm resistors, which all stay inside a +/-0.5ppm window over the last 2 years, so the drift of this group is also < 1ppm/yr.
As the 40K resistor inside the 3458A is not temperature stabilized, it can't be taken as an additional artefact reference. Anyhow, the correction factor (reading of 3458A relative to the median of the group) is inside a +1.8ppm / -0.5ppm window over the last 4 years.
My 3458A already has a 40kOhm, VHP101, which should also drift as low as the 5 other VHP202Z (typ. 2ppm/6 yr.).
The low variation of this correction factor evidences such a low drift rate.
Frank