There was such a thing as a carbon pile rheostat, or something like that. The stack-of-discs version of the carbon granule microphone, rated for more power and probably more stability, and controlled by a clamp rather than by a diaphragm.
For practical purposes today, I would suggest, I think, using the maximum acceptable value and adjusting the voltage reading across it with a larger value pot (or better yet, designing the surrounding circuit to have a controlled reference or gain); or if it's used as a resistor in and of itself, not for sensing or whatever, then perhaps a "power DAC" is acceptable (a few resistors, switched by MOSFETs), until the first method becomes applicable. (Example: a dynamic range of orders of magnitude, like for a SMU; the ranges could be switched by resistor selection, then each individual range is electronically variable.)
Tim