The 98-year open interval since the most recent ground-rupturing earthquake in the greater San Andreas boundary fault system would not be predicted by the quasi-periodic recurrence statistics from paleoseismic data. We examine whether the current hiatus could be explained by uncertainties in earthquake dating. Using seven independent paleoseismic records, 100 year intervals may have occurred circa 1150, 1400, and 1700 AD, but they occur in a third or less of sample records drawn at random. A second method sampling from dates conditioned on the existence of a gap of varying length suggests century-long gaps occur 3-10% of the time. A combined record with more sites would lead to lower probabilities. Systematic data over-interpretation is considered an unlikely explanation. Instead some form of non-stationary behaviour seems required, perhaps through long-range fault interaction. Earthquake occurrence since 1000 AD is not inconsistent with long-term cyclicity suggested from long runs of earthquake simulators.