I suppose one could have an system that drives downscale on a detected RTD open circuit, So it might read OK at lower temp, but the thermal expansion causes it to go open circuit and the fault mode drives the indicated signal downscale.
One reported failure mode is shorted element coil winding producing an overall lower resistance element. But it would have a low offset, not a reversed temperature coefficient.
An increase in resistance (loose or corroded connection) in the 3rd wire of a 3 wire RTD circuit that does lead wire compensation drives the indicated temperature lower.
A water-flooded field junction box could conceivably reduce the excitation current from the AI circuit by losing some current through a ground loop fault. Highly theoretical circumstance, given how few direct connected RTD's are wired through field junction boxes.
Or it might not be an RTD, it might be some other temperature sensor.
Thermocouples connected "backwards" (wrong polarity) drive the indicated temperature in the wrong direction.
A typical Pt100 RTD has a positive temperature coefficient. A thermistor with a negative temp coefficient connected to a circuit expecting a positive coefficient sensor will drive the indicated temperature in the wrong direction.
Usually miswiring a 3 or 4 wire Pt100 or Pt1000 to its appropriate analog input will cause it to report an offscale value or a fault code.