It's not the crash that invalidates it, it's the extreme variance. At one point, the S&P had two swings of 10% in one day. These kinds of swings were common, and they often happened without any shocking news arriving in between the swings.
Since the mid-1980's, stocks have traded as collectibles, not as cash flows. Thus the price of stocks is not based on fundamentals, but on game theory. The stock market is a coordination game with massive feedback loops in it.
In October 2008 there were some pretty shocking news and the market went down. I don't see how that invalidates the hypothesis.
EMH might have its weaknesses, and it might be even seriously flawed. But it's not invalidated because there was a crash or because people do "irrational" things, unless this irrationality is predictable.
(I believe the definition of "news" must be a bit wide for the hypothesis to work. If a big investor changes his mind about his risk tolerance and decides to sell a big chunk of stock, this is "news".)