Many economic experts have "seen" the crisis coming. Sometimes it looks to me, like a I-have-seen-it contagion.
Yes, there were signs in the fear index (the VIX) and leptokurtic distributions in special returns viewed a ruinous swarm behavior - Starling Flocks Are Flying Avalanches.
In this discussion questions, like Do Quants Destroy? were asked and Blame Me? BlameYou? Blame Math?
As pointed mathematicians, we also see threatening anomalies, when looking into deal type constructions, models, methods and implementations. When correlations really matter and we think, their estimation would take data back to the big bang, the alarm bells ring and if we find that in calibration the fit is so well, but the prices are so bad, as well. Even more, if we do deal type-across-model testing and fair prices are really different.
Yes we agree, lower model complexity is often better, especially if additional information gets lost in the numerical jungle - When Good ENUF Is Great.
But there are cases, where problem-complexity cannot be smoothed away. Especially in risk management, when in millionfold interplay of valuation errors multiply.
So, provided the world will not be restricted to a Vanilla economy, consequently, quant finance will be still required, also provided that the absolute of prices and inverting become more important, ... we sing This Is A Math's World.
Asymptotic Math methods, Finite Elements with Streamline Diffusion, Regularization techniques in calibration, Different Error Functionals testing in the calibration of stochastic volatility models, clever Least Square Montecarlo techniques ... that are accurate, robust and fast will really matter.