John....put you on the spot: if you were on a panel to determine conditional probability of San Andreas ruptures for different parts over the next 30 years, would you have the San Andreas fault from Golden Gate to 100 km north as higher than Coachella? (taking into account the new ideas you have brought up). These probabilities are probably a big deal...insurance rates may be set on them. I've always said, even before the Weldon/Fumal and Mojave work, that probabilities should not be too low or too high on any segment (Northern San Andreas and Carrizo were too low, others probably too high in 1988 or 1990. But the logic that quakes cluster you cite would say that San Francisco would have a higher probability than Coachella because it broke more recently. I don't believe this: Coachella is more likely from a structural geologic point of view. I don't think it has been demonstrated that strike-slip faults store much shear stress stress on segments completely ruptured...I think a paper by a (Mark?) Zoback showed that Loma Prieta had such a complete stress release and had left-lateral aftershocks on planes parallel to San Andreas.
Chris