Detection rates for 22 GHz water masers from O-rich circumstellar shells are always less than 100%. This can in part be understood by revisiting high angular-resolution observations of masers in OH 39.7 + 1.5 and red supergiants. The pattern of occurrence for water masers around the pulsation cycle of OH 39.7 + 1.5 suggests that a detectable maser only occurs when it is triggered by a density enhancement with a suitable gain orientation to our line of sight. Enhancements, in effect, break the local symmetry of many otherwise equivalent gain paths to provide a means for focusing emission into a narrow range of rays, which promotes a detectable maser. A brief review of masers in the circumstellar shells of red supergiants shows that water masers exhibit a rich ellipsoidal distribution of emission hot spots. Nevertheless, these usually contain no water masers on the axis of their biconical, higher velocity, lower density, bipolar flows. If many Mira variables and OH/IR stars have important bipolar flows, some will be seen nearly pole-on, and these are less likely to exhibit water masers, even if they host OH masers. The main input of energy (and momentum) to gas in a circumstellar shell comes from friction with dust particles driven through it by radiation pressure. In consequence, the switch between radial-gain masers in OH 39.7 + 1.5 that predominate near maximum light and a tangential-gain maser that can occur near minimum may result entirely from the pulsation-dependent temperature modulation of the shell. However, the fast polar wind in red supergiants is already well developed in the water masing zone, so its causality is probably located within the star itself.