The paper referred to uses momentum as a cross sectional strategy. The problem with trend following individual stocks within the stock market is the high correlation of all stocks (about 50% of a stock's return is due to the market factor and perhaps another 25% is a sector factor). If you short stocks you are vulnerable to sharp reversals.

I have been using momentum strategies for market timing - ie on the market as a whole not individual stocks - with good success. If you use it to short stocks aggressively you will get very volatile and skewed returns but using momentum of the market as a whole reduces volatility and improves the Sharpe ratio considerably.

Schleifer's paper "The Limits of Arbitrage" http://ms.mcmaster.ca/~grasselli/ShleiferVishny97.pdf gives some insight into why these things aren't arbitraged away. And the EMH is true only to the extend that arbitrage is possible.