It isn't the polls that are diverging so much as the poll methodology. If you look at the RCP spread this morning, you see six polls. Four of them have Obama up by 4, 5, 6 and 7 % respectively. Two, Rasmussen and USA Today, give McCain the edge by 1 and 4. BUT, the latter two polls are polls of "likely voters." These outcomes are therefore highly dependent on the way in which the pollsters determine who is a likely voter, whether they handicap or weight voter pools by likelihood to vote (a much better representation than simply declaring someone likely or unlikely to vote), and by the added uncertainly of any determination of likely voter at this stage. In short, Rasmussen and USA Today are garbage. Garbage in, garbage out. The polls of registered voters, about as clear a reading as you are going to get right now, show a pretty decent preference for Obama.
That said, I still think Obama has been ceding ground and/or opportunity for gains by allowing McCain to smear him and by focusing on foreign affairs rather than the economy. "It's the economy, stupid!" My prediction is that Obama is ultimately going to win in something approaching a landslide, although it will happen suddenly near the end after seeming close for almost the entire campaign. I base this primarily on the condition the economy will be in in October. But Obama had better do the job between now and then of framing McCain and keeping everyone's attention on economics. Than nature will take its course and the Republicans will receive a generational thrashing. A 60-vote majority in the Senate is possible, not likely but possible.