There are some physical limitations with respect to increasing torque output of a motor (AC or DC). Basically, magnetic flux=torque in any motor. The factors affecting flux are the current (ampere-turns) and the magnetic circuit formed by the "iron" of the motor in question. For DC, decreasing the resistance of the circuit (field or armature) by rewinding will increase the current (and flux). However, the result is not linear because the iron will become magnetically saturated at some point (true for any motor).
With respect to armatures, the amount of iron present in a design is normally the minimum amount due primarily to inertia considerations. For field frames, it is limited by frame type and economics (ie. once you have enough iron for a given torque output in a certain frame type, more iron costs more money with no benefit).
Another consideration is purely physical. Assume that iron saturation is not an issue. For fields, there may (or may not) be enough physical space to accomodate a larger wire size. For armatures, the slot size is almost always no larger than that required to accomodate the necessary wire and insulation for the original design. As well, the commutator riser is sized for the designed wire. Sometimes you can get away with larger wire, but usually it is a compromise (if it is possible at all).
Next, specific to DC, the brush grade will need to be changed to maintain proper commutation with the increased armature current. Also, (with any motor), cooling will be an issue.
From a performance standpoint, increasing torque by increasing field current will decrease the rpm at original rated armature current. Increasing armature current will increase rpm at original rated field current. To change the rating of the motor while maintaining nameplate rpm will require a proportional increase in both.