As a rule of thumb, the limit of the technique is about 1/10 of the pixel size in terms of the horizontal displacement you can measure. So in principle, you could measure horizontal displacement with an accuracy around 20cm using SPOT5 2.5m images, assuming you possess a good DEM and that all stereoscopic effects have been properly compensated for during orthorectification (or that you have access to images with very similar incidence angles).
If the fault creeps at 2.2 cm/year, then you'll need images separated by at least 10 years to properly detect it.
The same idea applies with SPOT4 10m images. 1m offset is the limit you can measure. It is possible in some ideal cases (not decorrelation at all, no vegetation or changes in the surface, images with close incidence angles).
In any case, the deformation you're trying to measure should be well localized and should not be too distributed. The technique works best when imaging high gradients of deformation. If the gradient of deformation is low, InSAR is usually more appropriate.
Hope this helps,