As the Economist?s Buttonwood points out, when everyone heads for the exit, all get squeezed. This defeats all hedging strategies. The late eighties and nineties brought us the end of a formulaic approach to politics (Socialism vs. Capitalism); the current decade follows up with the end of a formulaic approach to investing (Long vs. Short, Stocks vs. Bonds etc.).
A brilliant comment by Hyun-U Sohn sums it up:
The article betrays the view of correlations and other statistics as describing laws of some kind. That you distinguish between dynamic and static correlations and between "short"- and "long"-term ones doesn't really make a difference. The implicit belief that there are hidden laws, to be uncovered by data analysis and then extrapolated to future investment decisions/ advice, is one of the more fundamental, philosophical factors behind today's crisis and the rise of financial "engineering" that preceded it.
My own view is that there are no laws, only analogies. Statistics can serve as summaries of history and as expressions of subjective beliefs about the future or about hidden facts. The circumstances that led to the data showing low correlations between, say, commodities and equities in one period were specific to their time. They may or may not look similar in the future, and even if they were completely dissimilar, they may still give rise to a set of similar data for a while before the dissimilarities show, perhaps abruptly.