I'm surprised at the lack of response to this article and votes for it over the last year. This is pretty important for optimization. I think the best advice from this article is: "If the column is not needed then don't include it." I'm guilty of the "just in case I need it" method myself. I will certainly be limiting the columns to only the necessary ones in the future after reading this article.

The same concern that Journeyman posted also worries me. I worked with a database that was indexed to death and the indexes made things worse in some cases. I think I would save moving to a non-clustered index or a covering index as a last resort. (That is if the table has an index to begin with.)