I have thought about it as well. The signals that could determine relevancy in your browser history are likely different form the ones that determine it for web search, and measuring engagement with a page is a good example (it would help web search as well, but it would be harder to measure).
Other things you'd take into account are:
* continuity in time - reading several pages on similar topics one after the other.
* analysis of the subgraph of the web graph that you visited
* you'd be more aggressive with query expansion because you work with much smaller corpus, e.g. you'd want to take hypernyms and hyponyms into account
Ah yes, engagement is the word. Interesting. I'd imagine the query expansion should also work across the author dimension--include and rank matches from the same blog higher, even if according to my history, I haven't read that post.
Other things you'd take into account are:
* continuity in time - reading several pages on similar topics one after the other.
* analysis of the subgraph of the web graph that you visited
* you'd be more aggressive with query expansion because you work with much smaller corpus, e.g. you'd want to take hypernyms and hyponyms into account