Martin Herink has an interesting piece (here) at Gamasutra about the "problem of the cutscene," a problem which he's seeking to redefine: instead of debating whether cut-scenes should stay or go, Herink argues, we should be talking about how they work from a formal point of view. Designers should think about certain cinematic aspects of cutscenes when they produce them, and critics should think about those same aspects of film-language when interpreting the games they're part of.
The piece is couched in language that's too self-consciously scholarly for my taste, but if you're interested in the sandbox-to-rails continuum (like me, here), it's definitely worth a read. Certainly Herink is right that cutscenes aren't the enemy; rather their use and production needs to develop in the same way that film-language developed (cf. Michael Abbott's great post on Griffith and Kojima, here), just like the rest of the art of the video game. I suspect that as that art moves forward, cutscenes will dwindle in number, but grow in significance.
One of Herink's most interesting points is that cutscenes are by nature contemplative, while "regular," interactive gaming is kinetic. I don't entirely agree that the distinction makes sense when framed that way, since I believe in a continuum between the two. On the other hand, I think there's a really interesting comparison to be made with the way certain sections of ancient epic (a feast scene, for instance) would have been almost entirely the same from bardic occasion to bardic occasion (like cutscenes), while others would have involved a great deal of improvisation upon and interaction with the existing story on the part of the bard, and with him his audience.