Wow. I must say your original design didn't really excite me too much, but after all these Rev. C design decisions I'm pretty damn excited. I thought every single thing you changed was for the better and now I'd actually buy one.
Since you asked, I did have a few comments
1) Why not move the battery measurement divider to the 'system' side (pin 3) of the switch rather than the 'battery' side (pin 2) of the switch to save the quiescent current when the supply is turned off? Not that it's really a big deal, if your battery is about 3AH I back of the envelope around 2 years to discharge the batteries into that 44k. Doesn't cost anything to change either though.
2) You've probably considered this, but at 20V out and ~7 Volts nominal on the battery string to get 1A out of the output you'll need close to 3A at the batteries. Can your cells handle that? You mentioned they were about 3AH capacity so 1.5AH each? That would be a 2C discharge rate? Sounds a little quick to me. Maybe a good candidate for a firmware fix, just drop the maximum current limit for high voltages.
3) Looks like the 100pF compensation capacitor from Vo to the feedback pin on the switcher is missing from your schematic. I also notice that the digital pot seems to have around 16pF of output capacitance. I haven't gone through the math but could you be introducing a stability issue? The series 820 ohms helps, but 16pF seems close enough to that 100pF to be worth a second look. Maybe another argument for some kind of analog control loop for the pre-regulator?
One last note: THANK YOU for using Ethernet comms and not USB/GPIB. (Why isn't GPIB dead yet?) Ethernet is perfect for test equipment: isolated, completely ubiquitous, run long lengths with no problem etc. But it seems like all sorts of gear just still targets USB or GPIB and not Ethernet and I just don't get it.