HST provides an unparalleled venue for high contrast imaging which has enabled new observational domains in exo-planet and debris disk imaging. Unburdened by atmospheric 'seeing', NICMOS and STIS achieve very low levels of background contamination from the wings of stellar point spread functions (PSF). Coronagraphy provides additional contrast gains approaching an order of magnitude at small angular distances from occulted stars. The stability of the platform allows scattered and diffracted light to be further reduced by two additional orders of magnitude through PSF-subtraction. The non-destructive read-out modes of the NICMOS detectors permit sampling the PSF, with its strong radial brightness gradient, over a dynamic range exceeding 5x107 in a single spacecraft orbit. In H-band, sub-stellar companions of ΔH ~ 8+2 x (angular separation in arcseconds) are unambiguously detected in twenty minutes of integration. Raw sensitivity metrics, such as presumtively static Strehl ratios, are often invoked in comparing the performance of different instrumental systems but belie the true detectability levels which are dominated by systemic non-repeatable PSF variations (not photon statistics). Such variations can give rise to false detections of companions (and circumstellar disks) and introduce very significant photometric errors. The ability to rotate the HST field with high precision about the target axis and acquire temporally stable reference PSFs readily permits the identification and rejection of rotationally-invariant optical artifacts. We discuss the repeatable, quantifiable performance limits routinely reached by HST (currently unachievable on ground-based systems), for which PSF stability is critical.