Rough approximation.
As i remember.. at the time of the use of 10kOhm to 1 Ohm transfer presented on diagram, DFM used Guildline 6675 DCC (Direct Current Comparator).
6675 Comparator has Accuracy <0,1 ppm/year/3degC for 1Ohm...1kOhm, and <0.15ppm/year/3deg C for 1kOhm..10kOhm. Stability is accordingly 0.03ppm/day/1degC and 0.04ppm/day/1degC..
Guildline 9350 transfer has accuracy about 0.05ppm
Comparing 10kOhm to 10kOhm, transfer from 10k to 100 Ohm, comparing two 100 Ohm, transfer to 1 Ohm, and comparison 1 Ohm to 1 Ohm will give
sqrt(,15^2+0,05^2+0,1^2+0,05^2+0,1^2)=0,22ppm.
This is upper limit of uncertainty of transfer from 10kOhm to 1 Ohm, because real uncertainties are lower than specified above.
There are also other components of uncertainty budget, but they are usually smaller than this two main components, taken into account.
Stability of very good SR104 is about 0.05ppm/year, (I think DFM has this kind), so the group of 3 resistors should lower this to about 0.017ppm. This woud be the starting point.