Don,

When thinking about the break in a big earthquake, just take the simple view that the two sides of the fault are trying to move past each, but are stuck (as you know, I'm sure). The deeper region below the seismogenic limit, about 10-20 km in California, is just slowly flowing.

In a big earthquake, the entire seismogenic depth range break, and the deeper area has accelerated flow for a while due to the load imposed by the upper part sliding.

As Chris alludes to, some parts of the fault plane seem to stick more than other. When a particular part of the fault slip more than elsewhere in a quake, we call it an asperity. There are arguments, but we think asperities persist across many earthquake cycles.

Foreshocks may be caused by yielding of weaker parts of the fault, which then add extra stress to the locked asperity, causing it to fail in turn. Or foreshocks may just be linked in a cascade with mainshocks the same way mainshocks cause aftershocks, only in the former case the later earthquakes are bigger rather than smaller than the earlier ones.

I think you're bringing up the long-term evolution of the geometry of the fault, which must happen, but isn't linked into our picture of foreshocks and mainshocks. The geometry does not change significantly in a single earthquake, although it may be that the fault is weaker (or stronger) where the geometry is complex and slowly changing as opposed to where the fault is simpler.

Not sure if this makes sense, and I didn't proof-read it.

John