This misses out on what the Google "cruft" is all about. Products like Android or Chrome aren't meant to be standalone products/business (and wouldn't be profitable if they were). Their nature is a defensive moat around Googles Search/Ad business.
However, I wouldn't go so far as to say the article's idea is an all around bad one. If the described sister companies were how new ideas got brought to the Google table, and only after their success were they integrated into the moat, then Google can persist with their current business model (Castle-Moat), and the rest of the world can get sans-barnacle developed products. I think this might actually fit into that entire "returning to startup roots" initiative that Larry Page keeps mentioning. If you think about it this moves all the development overhead mentioned to the integration with Google rather than the core feature development of the product, so no good idea gets hindered by red tape, and integration can then thereafter be prioritized by the degree of each products success.
I realize I'm talking abstractly, from a far, an without any account for constants, project nuances, or implementation details, but I wanted to post about the techcrunch moat/castle article, and when I saw you had already posted it, decided to give a brain-dump instead. Enjoy! haha...
To add on to the this, the author also ignores the many benefits of keeping a variety of products under one organization. By keeping the products within a company, you enable much greater levels of collaboration than if they were separated. For example, Google can share private APIs between Gmail and Buzz, however, if they were separate companies, all shared APIs would be legally required to be publicly documented - even if the companies were buddy-buddy. It is also much safer from a disclosure perspective when you want to coordinate projects to minimize the time delay between the release of products that integrate with each other.
Sure, there is overhead introduced, but that is exactly why companies like Google hire PMs. Their job is to manage the coordination and minimize the overhead so that developers can actually write code.
I think what he's trying to get across here is that there is a significant amount of overhead that is getting in the way of Google's own culture. Breaking it up into a conglomerate of smaller pieces would typically _increase_ this cruft -- Microsoft even in its current state is a good example of this. At Google however, this would _optimize_ the system instead, making it leaner, meaner, and as passionate as the employees it is losing.
(The blog post doesn't mention it clearly, but the author is an ex-Googler)
I wouldn't call it broken, especially compared to Microsoft (I've worked for both). Besides, why don't we at least give Larry a few weeks as CEO before proposing a restructuring :-P
Relevant article: http://techcrunch.com/2011/03/25/search-googles-castle-moat/