Between 1995 and 2009, electron temperature (Te) measurements of more than 15000 plasmas produced in the Joint European Torus (JET) have been carefully reviewed using the two main diagnostics available over this time period: Michelson interferometer and Thomson scattering systems. Long term stability of JET Te is experimentally observed by defining the ECE TS ratio as the ratio of central Te measured by MICH and LIDAR. This paper, based on a careful review of Te measurement from 15 years of JET plasmas, concludes that JET Te exhibits a 15-20% effective uncertainty mostly made of large-scale temporal drifts, and an overall uncertainty of 16-22%. Variations of 18 plasma parameters are checked in another data set, made of a reference data set made of ohmic pulses as similar as possible between 1998 and 2009. Time drifts of ECE TS ratios appear to be mostly disconnected from the variations observed on these 18 plasma parameters, except for the very low amplitude variations of B0 which are well correlated with off-plasma variations of a 8-channel integrator module used for measuring many magnetic signals from JET. From mid-2002 to 2009, temporal drifts of ECE TS ratios are regarded as calibration drifts possibly caused by unexpected sensitivity to unknown parameters; the external temperature on JET site might be the best parameter suspected so far. Off-plasma monitoring of MICH made of calibration performed in the laboratory are reported and do not appear to be clearly correlated with drifts of ECE TS ratio and variations of KC1D integrators. Comparison of estimations of plasma thermal energy for purely Ohmic and NBI-only plasmas does not provide any definite information on the accuracy of MICH or LIDAR measurements. Solutions aiming at tracking down these unexpected uncertainties of JET Te are detailed and can be performed during next JET campaigns (C28+, after October 2011), for instance with highly-reproducible reference pulses and off -plasma monitoring of the diagnostics. Whatever causes these Te drifts, this experimental issue is regarded as crucial for JET data quality.