If turbine blades (is windmill too old fashioned?) didn't follow the wind around, the problem would be more containable. A turbine pointing axially at the RADAR head would return negligible doppler shift whereas one presented at 90 deg would return the maximum for the rotation rate.
Depending on spacing, do blades reflect and return from each other, thus multiplying shift values?