If it were my site, I'd add a document classifier that makes articles about certain topics face a headwind, or better yet, creates a kind of repulsion between recent articles... If there are already two articles on the front page about Assange on a given day, that's enough.

If documents are clean, you can compute the repulsion from the bag-of-word dot products between documents, but web pages often have a lot of noise in the outer template, so getting good results with web pages means you need some kind of template drop-out.