If you look in a catalog for brushless PM (servo) motors, and analyze the torque-vs-speed capabilility curves, you will see on many plots that the continuous torque rating gradually diminishes in a linear or approximately linear fashion as the speed increases from zero. This is for speeds below those that are voltage-limited.
If the copper resistive losses were the only losses, the continuous (i.e. thermally limited) torque rating would be constant up until voltage-limited speeds. This rating is based on the idea that the motor is able to dissipate a certain number of watts continually (but read the fine print for the conditions under which it can do this).
The downward slope of this rating as speed increases therefore represents the watts/krpm lost due to other mechanisms. To the extent this curve is really linear, this value is a constant. What I don't know is how well motor manufacturers really characterize this relationship, or if they just draw a straight line between the zero-speed case and a high-speed case -- which would be a conservative approximation.
Curt Wilson
Delta Tau Data Systems