I have it sort of working in one scenario. It's definitely dependent on friction, and it requires the object to be moved to have a constant force orthogonal to the surface with the constant_linear_velocity value. The orthogonality of that force seems proportional to the velocity provided. If I introduce a perpendicular force, or if I rotate the angle of the surface, the effect is proportionally weakened.
Perhaps this has primarily to do with how friction is being calculated, and friction is clearly a primary contributor when using constant_linear_velocity. Unfortunately, It's a major pain in my ***. Gravity is a constant force with a vector opposite the "lift" I want from constant_linear_velocity, and gravity also reduces the orthogonality of forces on my moving object to the surface. Furthermore, if my surface is rotated around the Y axis, the the force pushing my object into the surface is even further askew, making the lift even weaker. All I want is a constant, consistent upward Y velocity applied, but unfortunately it appears this is not a way for me to achieve that.
EDIT: Actually, I just noticed that the friction coefficient can be cranked higher than 1, and this seems to help a lot. I'll toy with this...