Perdix,
As long as you can null out all the pots, you're unit -should- be back to operating condition and in calibration accurate to +-0.1ppm. Unable to null S2 is a common drift problem but if that even gets to within 5uV of null you've got it very close to dialed in. You should get a realistic sanity check on your 3458a - just keep your 3458a locked in to 10V range (if you're testing your 720a by dividing a 732a/b) and remember that any error you see will be relative to 10V - or whatever your input voltage is. The left 3 dials will be most important, the other 4 are usually close to perfect anyway if the switches are clean - unless the unit has been overloaded. Remember you are checking for a divided ratio out of the 720a, so keep cross-checking your 3458a between the 720a actual input vs. actual output. Low thermals are a must here also.
Also check extended settings such as 0.5000000 vs 0.499999X - They should be very close.
The problem with sending a 720a back to Fluke for repair is hardly anyone at Fluke knows how to repair one of these. Not efficiently anyway. They tend to drop in whole switch assemblies when a single resistor might have drifted.
Anyway, good luck!. These are a fantastic tool to own if you're interested in real accuracy. If you get it to complete the calibration procedure you've just saved yourself $46,255 over the cost of a new unit. So it pays to spend some time going through the cal sequence carefully and completely - and keep it in tip-top condition.