Joel Spolsky and Dmitri Zimine mixed it up last November over whether or not to interrupt a team in the middle of an iteration. I'm pretty late to this thread. But, I think I have something to add. So, here goes...
Dmitri started the thread by arguing on behalf of never interrupting an iteration once it starts. Basically, he's just parroting the party line from the Scrum doctrine: Management is not allowed to interfere with the operation of a team during a sprint. If they do, the team can cancel the sprint.
Joel countered that you should consider the nature of the interruption before blindly preventing it. He argues in his inimitable way that the financial independence of the individuals involved could very well be at stake. He may as well have been yelling: "Take door number 3!"
What both guys missed is that they are arguing the opposing sides of a pendulum, and that the real truth probably lies somewhere in the middle. (Ironic considering the tagline on Dmitri's blog.)
Here's my advice when faced with the decision of whether or not to accept new work after an iteration has started:
There, now. Everyone has visibility into the process. And, the person in the best position to make the call as to priorities - the one who knows both the benefits and (now) the costs of swapping in new work - is the one responsible for making the decision. That's what I call Agile Prioritization.
For further study, I recommend that Dmitri and Joel (and everyone else) read Mike Cohn's excellent book: Agile Estimating and Planning.