Well, assuming a cable that is a standard 3-wire + shield mic cable, used unbalanced and with the "cold" wire connected to the DC bias on one end, this would create a DC "bias" field in the dielectric, wouldn't it? And the signal wire would simply overlay that with AC field one or two orders lower in magnitude, in terms of polarizing power.
In a true triax with the inner shield applied to the bias voltage this would work equally well if not better.. still under the assumption that the dielectric losses are indeed not perfectly linear and time-invariant in some way at very low signal levels, of course... OTOH, in high impedance sensor precision interfacing dielectric loss is definitely an issue, and when it is non-linear it will introduce distortion. I personally would not dismiss the whole concept, though generally speaking I agree this is a non-issue for the most part, for high-level signals driven from standard low-Z (100Ohms or less) outputs.