Sensor drift versus a loose ferrule
A food dryer showed a slow climb in reported air temperature over three months. The product was not scorching, which already argued against a true process rise. The local files looked like textbook drift.
On the visit the RTD well was finger-loose. A fitter had cracked it during a gasket change and closed it ‘enough’. The probe sat a few millimetres out of the airstream. As vibration worked it further out, the reading climbed toward cabinet ambient.
A quarter-turn and a paint mark on the nut ended the trend. We still logged a calibration check against a reference probe, because a loose well and a tired element can coexist. They did not in this case.
Before you discard a probe for drift, put a wrench on the well and a reference next to the tip. Ageing happens. So do quarter-turns.