I agree that stability is performance over the life of the instrument, and also closely related to predictability. One example may be that I take the instrument into calibration and determine it is within limits but has drifted downwards and needs to be tweaked. If over many times of performing calibrations, and it is always drifting downwards between cals, I might invoke that the instrument is not "stable". A good way to identify stability is to plot calibration results over time on a control chart.