Nice article, I especially near the end where real world experience is discussed.  I personally would like to see real world comparison between nolock and readpast, but maybe that's another article.  

In some places the article correctly attributes decision-making to the optimizer, but in a few places I think the article misleads by attributing optimization decisions to the Query Analyzer:  "Table Hint May Not Be Used By Query Analyzer", "the query analyzer will often choose ...", "the query analyzer's preference for indexed views...", and "Another reason the query analyzer may ignore...."  Substitute "optimizer" for "Query Analyzer" and problem goes away.