Brad DeLong expounds a switching model to explain some of the asymmetric negative moves involved in crashes.
I've long thought that the next step in financial models was a move toward adaptive model making. In other words, people look around them to see what works.
Your model for the weather is either a) look out the window, b) today's weather will be like yesterday's or c) whatever the weatherman says. For financial markets, sometimes it's smart to act like everyone's rational and pricing just on market beta, sometimes you need your Fama-French, sometimes you need temporary state variables or non-linear factors, and for several years in the late 90's it made sense to just invest in Internet stocks.