Thanks to both of you for the insight and encouragement. I dusted off some of the old physics books last night and started to do the calculations to really find out where the limitations were. The conversion factors were a bear, so I did a little more searching online and found
http://www.linearmotioneering.com/
W00t! all the hard stuff is computed.
From what I understand now, a 100lb x axis with a 5tpi screw, 120ipm rapid, .1m/s2 acceleration and direct drive will require 80Ncm @ 600rpm
The stepper motor torque curves from Keling show that there are NEMA23s that will do this. The big problem is the mill head because not only do you have inertia, you have to overcome gravity. Here you would need roughly 1Nm @ 600rpm. That is NEMA 34 territory, but adding gas struts or counterbalance could reduce the torque needs.
So, the main reason to go to a larger motor is acceleration and faster rapids. The faster you want to get the axis moving, the more oomph you need from the motor. And since the steppers drop off in torque very quickly, you run out of oomph to keep accelerating.
Can these limitations be overcome? I think so... I'm looking at using Mach 3 and I think that it has acceleration settings. If I limited the acceleration or maybe used some of the anti-jerk settings in Quantum, the motor system may not have to be as over engineered as it was in the past.
Hmm, cost of tinkering time + risk of being wrong + cost of head strut parts is > the ~$500 premium to go to NEMA34 setup