Here are some initial observations on this thoughtful post from the AHA blog:
Many times it is the more specialized journals that contain the articles that most interest a researcher. The AHR covers a broad range of topics, and rightly so, but often doesn't have the more narrow source material that needs to be cited.
It would be interesting to track links going bad just over the last few years given that that appears to be less of a problem than in the past. Standardization of I.T. system upgrades that change website configurations and utilizing bots to track those updates down are two possible solutions to "lost" URLs. A significant number of our peers still dismiss the web as a big black hole instead of seeing its true potential - especially for those of us without major research libraries readily available geographically.
Additionally, search engines like Google have become much more useful in finding matches to exact titles if for some reason the original link goes bad, especially if the correct metadata has been applied to the document before placing it online.
It will certainly be interesting to see where we are even two years from now. Tools like Zotero certainly make research much more manageable.