While I might color the Canonical-Intel relationship a bit differently, generally, you're right that a collaboration of this magnitude between completely different companies is very risky. Unless something very special is invented to make the planning and integration of the work smooth, in my mind, a project like this will naturally fail.
"Something very special" might include a unique way of setting goals and measuring results, or maybe setting up a joint office and relocating everybody, or coming up with a new, groundbreaking continuous integration system, or hiring the most stellar open source team in existence and giving them good leadership. Probably all of the above and more.
It just seems that the decision-makers at the companies involved didn't really want to try something different.