Nate Silver wrote a great article in 2010 about why the notion of exit polling itself has a lot of fundamental problems[1]-- among them, the fact that it is extraordinarily difficult to get an truly random sample.
If you read the article, there are several instances of the same problems causing very large errors in elections prior to 2000, for example, several miscalled states for Clinton in previous elections. To claim that these polls "suddenly" stopped working is demonstrably false.
[1]http://www.fivethirtyeight.com/2008/11/ten-reasons-why-you-s...