Steve, while I've posted that I would not expect a M9 on a thrust fault that just had a M9 for many years (not sure how long "many" is), in the case of Parkfield, you might want to consider the slip distribution of the M6.0 2004 quake. There may be large parts of this rupture that did not slip very much and thus it might not take too long to build a similar amount of accumulated strain (8 years * 3.5 cm/yr =24 cm).

More controversial, but published by Weldon, Fumal, and others, is that you don't have to reload faults: that they can store strain for multiple quakes; not all the deviatoric/differential stress (I forget the difference) is released in a quake. John Vidale has commented on this here a few years ago, but I thought he went too far in the direction of the next quake being in same area as the past one (yes, I know about static stress changes).

Multiple quakes without reloading is bad news for short term pacredicition and maybe 30 year forecasting. Decades ago I saw a talk by Ralph Achuleta at UCSB tearing down the characteristic earthquake concept.

Chris