Assuming suitable degradation models and condition monitoring are available, the remaining useful life (RUL) of technical devices (e.g. switch engines, signals) or critical track defects (e.g. track settlement/misalignment, rail roughness) of rail segments under study can be prognosticated. Subsequently, ongoing maintenance actions can be optimally rescheduled. The prognosticated infrastructure performance indices, however, might be affected by various uncertain factors, as: (i) imprecise parameters of the applied degradation model (e.g. soil/rail quality and soil moisture), (ii) vaguely known operating conditions (e.g. future temperature and load stresses/tonnages), and (iii) empirically chosen tuning/nuisance parameters of pre-and post-processing steps (e.g. noise filtering and feature selection calculations). In practice, the impact of these uncertain factors onto the model result (e.g. RUL) varies strongly. Some of them can be changed by order of magnitude without any significant model response variation, whereas slight displacements of other quantities might have a serious impact. In sort, the model response can be sensitive or insensitive with regard to the mentioned uncertainties. The sensitive group of uncertain factors has to be treated with great care, the insensitive group, however, can be considered as deterministic identities and kept at fixed values.