In a previous entry about constant change, the idea of a horizon of predictability was introduced. This concept, along with the agile discipline of amplifying learning, suggest a strategy for addressing problems in a project.
Shorten the length of the iterations you are using. Contract your “planning horizon”. The length of your iterations should be motivated by the horizon of predictability for your environment. If your project encounters trouble, particularly of the sort where it looks like you might not accomplish your commitments for an iteration, then shortening the length of iterations will enable you to resolve your problems.
First off, by shortening your iteration length, your opportunities for learning become more frequent.
Secondly, a contracted planning horizon will put you more firmly inside the horizon of predictability… and therefore there will be fewer unexpected changes (on the whole, not in any specific iteration).
A related article is The Pros and Cons of Short Iterations.