The impact of disruptions in JET is well-known not only with the carbon fiber composite (CFC) wall, but also with the metallic ITER-like wall (ILW). A disruption predictor, called APODIS, was developed and implemented for the JET real-time data network. This predictor uses seven plasma quantities (plasma current, mode lock amplitude, plasma internal inductance, plasma density, stored diamagnetic energy time derivative, radiated power and total input power) and it has been working during the ILW campaigns in JET. It has reached good results in terms of success rate, false alarm rate and prediction anticipation time. However, it is important to note that any signal could fail during any discharge. If an incorrect signal is used by APODIS, this can be an issue for the predictions. Therefore, the first purpose of this article is to determine the robustness of APODIS. Robustness is the predictor reliability when a signal fails. To determine the robustness, anomalous signals have been simulated and the quality of the APODIS predictions has been estimated. The results show that some signals, such as the mode lock and the plasma inductance, are essential for APODIS to provide a reasonable success rate. Under the failure of other signals, APODIS performance slightly decreases but remains acceptable. On the other hand, during the ILW campaigns, APODIS has missed some disruptions due to a lack of temporal resolution in the prediction. Owing to this reason, a second analysis has been carried out in this paper. The effect of increasing the prediction temporal resolution has been analyzed. The plasma signals are digitized at the same sampling frequency (1 ksample s−1) but a sliding window mechanism has been implemented to modify the prediction period from 32 to 1 ms.