The steering motor gearboxes are 50:1. They do spin very fast, so it would make sense to gear them a little lower to save some battery energy. We have not noticed the motors getting hot, so that is good.
The real test would be trying to steer while being pushed. I know that would take much more torque. We are using needle thrust bearings in the module pivot, so that should help somewhat.
Your model leaves out the friction force in the pivot bushing or bearing. You should also consider the extra toque that would be needed if the robot is getting pushed:
pushing force*height from pivot to ground*pivot bearing radius*pivot bearing coefficient of friction/pivot bearing radius/2= turning torque
N*m*m/m= N*m (unit sanity check)
The divide by 2 is assuming that you have 4 modules, but only 2 take the load because 2 are somewhat lifted by the pushing robot.
1640 did use a window motor for steering,
but they only did it for their first generation pivot drive. If you do use them make sure to remove the backdrive locking pins. There is info about that on their webpage, but I can't find it just now.