Lars - btw how did you calculate your RH coefficients / do your RH experiment?
About every second month I spend a couple of hours to measure a couple of hundred resistors (and many DIY 10V refs) against my HP3456A. I have done this for almost 20 years.
Lately I have also started to use ice-cream boxes and large bags of silica gel prepared for different humidity’s. Right now I have a VREF10 from Voltage standard under test.
See also
https://www.eevblog.com/forum/metrology/calibratory-d-105-dc-precision-voltage-reference-standard/msg803701/#msg803701
for an example of my use of the box that revealed a 20ppm change due to humidity on the 2ppm specified D105 ref.
For the long-term tests with the HP3456 I use regression in Excel to get temperature, humidity and aging. I enclose a couple of screen shoots from Excel with temperature compensation but not aging and humidity.
The first chart is 100ohm, 1k, 10k and 100kohm wire wound 8E16. The 1Mohm is ordinary 1% metal films I happened to have 20 years ago, if I remember correct I selected four 220kohm + some smaller to get to 1Mohm with low tempco. (As a side note I have seen up to 1000ppm/year drift on some new 0.1% metal films from Reichelt).
From 1997-2001 the values are relative to the 3456 and a small extra humidity sensitivity about 0.1-0.2ppm/C can be seen due to the 3456. From 2001 I got a GR1440-10k that is the reference and the 100ohm, 1k, 100k and 1M is related though a home built resistance decade. I enclose a picture of one of many home-built decades I have including my parallel and series-parallel adapters.
I also enclose a chart showing 100ohm, 1k, 2 pcs 10k and 100kohm VHP resistors I have had six years. Remember that the 3456 has 1ppm resolution and to get from the 10k GR1440 to for example the 1kohm I have to measure the GR, the VHP and series-parallell 10x1k+ series 10x1k+one of the 1k. So the ripple is not so bad, sometimes it also can be seen that I have some wrong values. Nowadays I have an automated routine that takes 20 samples with 10PLC but disturbances can still occur. Earlier I averaged a few samples with 100PLC in my head.
The last chart is two each of 100ohm, 1k and 10k of VSMP 1206SMD soldered on two small FR4 board.
Lars