The OSS community has typically favored the former, while denigrating MS for the latter.
There's benefit to the integration path. MS' tools play very nicely with one another, and it would be nice to see improvements on the OSS side.
However, I don't think the Gnome project should go that route. Part of its success has been its modularity - it has been long-lived, and used in unlikely places. Choosing the integration route significantly reduces future options. Heck, what if Oracle decides they don't want the expense of Solaris, and spin it off into its own company (yea, yea, I know)?
If Canonical (or other downstream) chose to do something like this, I would think it was a great decision. Ubuntu could end up with a superior experience, if done right.
I don't see the point of a core project doing this. There's nothing about the current model which prevents implementing this downstream, or as an add-on for those that are interested.