This deck, from a recent POPL presentation, offers a fascinating view on the future of mainstream games-programming languages, specifically around safety and concurrency: http://www.cs.princeton.edu/~dpw/popl/06/Tim-POPL.ppt.
(Of course, if I had a link to a video-taping of the talk or a detailed paper, that would be more useful; but I don't.)
An interesting conclusion--reinforcing a growing commonplace belief, at least here at Microsoft--is that there is no single concurrency model that works for all problems. In other words, there is no silver bullet. A combination of isolation (via STM), with implicit- and data-parallelism is discussed.