Moving all this into the new metrology folder has the nice effect that it is kind of reviving.
Anyway, a few comments:
SR104: I bought one around 2 years ago, it was built in the 70ties, to replace my home-brew 10k reference resistor (VPG hermetic, which has a track record so far to drift about 0,2ppm/a). At the time, I did not have a meter with a guaranteed ohms transfer accuracy (the 3458A does not have that). So I decided that I would apply a stable voltage to the SR104 in series with my 10k ref and compare the voltages (the 3458A is pretty good at that). That resulted in a deviation of less than 1ppm between the two, so by coincidence, they were more or less equal (at the 23C I measured them). I then, about 1+ year later, got other meters (a 7071, 1ppm resistance transfer accuracy, and a 1281, well below 1ppm transfer acc.), so I decided to redo the measurement, and in comparison also with a 3458A in ohms mode. To my surprise, I measured about 7ppm delta. I also redid the voltage drop method described above, essentially same result. That eliminated the dielectric absorption issue (known from the 3458A, using OCOMP, but I did use a delay anyway, and switching OCOMP off did not make considerable changes, and so I had not expect this to be the issue anyhow; see also Dr Franks story about this here). Conclusion was, something has drifted, rather than the current measurement being wrong, too many equal results with different meters. A SR104 does not drift 7ppm in about a year (actually, if the last measurement was right, it would have drifted less than 5ppm in 40 years). I had the suspicion that my SR104 maybe had lost its sealing, so I decided (yes, I know the comments that I will be getting) to look inside (I just lifted the lid a bit, so not much stress applied). It had no history, other than its original value, and it needs to be call'ed now anyway, so this was not a loss for me. And at the moment I cannot use it. No leak recognized though. So next third party cal. , due soon, needs to tell what drifted (it could of course as well be that my initial measurement was wrong for unclear reasons, we will see).
What surprised me was the wiring. This is supposed to be a 4-wire resistor, so each lead at the resistor terminals being connected by two wires to two binding posts. That way the wire resistance is eliminated. What they had done though was to connect the two binding posts of each side directly, and then connect two wires to that pair. That does not make any sense. I changed this, the change in resistance was marginal (0.6ppm less, one would expect a slightly lower resistance as the copper cable resistance is eliminated; ma hope was a bad solder joint too, explaining the 7ppm above when I saw the strange wiring, but that was not the case).
SR1010: This is normally used to transfer resistances, rather than as a reference. So normally, before their use, they are calibrated (important is their deviation). I bought a couple for them (10R, 1k, 100k, 10M; per resistor) with bars recently, for the purpose of transfering 10k to a range of 1 ohm to 100M with better accuracy than I do so far. So what I want to do in specific is going from the 1k (10 in series is 10k, 10 in parallel 100R) and measure each one with a (non-calibrated, at least not to that accuracy) meter in transfer mode. That gives certain deviations. Then, with 10 in series, I would compare this value to the precisely known 10k standard and transfer the total deviation measured to each of the single deviations I had determined before (relative, but not absolute). That way, although I do not have a 10k SR1010, I still can do a proper transfer from the 10k standard to the 1k SR1010. Al this with an error propagation calc should give a good transfer within the abilities of the SR1010 as per manual. Any objections to this?