So I was building a Silicon Chip EHT Stick (April 2010). It's just a 1000:1 passive divider. 80 10M resistors in series for the high side and an 820k + 30k + 50k Pot on the low side. The theory is you set it up against the 10M input impedance of your meter and adjust it for about the right output.
So I built the board up and thought "Well, my 3457a has a 3G range, let's measure it". Now to be fair, the 3G range is one of the "extended ohm" ranges and the stated accuracy is +/- 16%. Hardly worth it.
So with the board on the desk I put my trusty croc clips on and got nothing. The meter just read overload. Hrm, let's try the 4 wire mode. Nup. Nada.
So I thought since I have a PG506 with a vaguely accurate 100V output I'll set the divider up on that and use that to adjust things. That worked, For a 100.05V input I could get close enough to 0.10005V out. I measured the bottom end resistor chain and got about 868k . Put that into the spreadsheet in parallel with the 10M input and got ~798k. Calculated the top half out at roughly 798M which is pretty good for a string of 80 resistors.
While I had dinner I wondered about the interference induced in a set of 700mm test leads at the poofteenths of an amp you'd get through an 800M resistor so I hooked it up with my 100mm leads and hung it off the front of the meter. Bang, got a fairly steady reading at 100NPLC. Reading was ~7974**k. Within sneezing distance of my calculated 798M. Nice!
For future reference, when measuring more than about 20M, use the *short* leads!