It pains me to say it, but this particular ticket might rise to the level of requiring yet another immediate release http://trac.osgeo.org/postgis/ticket/2084 We're returning wrong (backwards) results for fairly common predicates (point/poly within and point/poly coveredby) P.