Take Office, for example, where it's a benefit to users (particularly a user who knows Word but doesn't know Excel so well) when the Office programs share many of the same controls, idioms, shortcuts and functions. Before the days of Outlook, Office users were very frustrated because the Exchange email client didn't feel like an Office application. The reason for that was clearly that the managers of the Exchange email client weren't the same as the managers of Office, so the teams couldn't work together and share ideas effectively.
In that case the problem was solved because Outlook was developed as a rival client by people much more closely tied into the Office software group management. Only that codebase developed different problems over time -- Outlook didn't work as seamlessly with the Exchange server, causing disruptions that frustrated users in different ways. Clearly this was because the Outlook managers had only distant links to the Exchange server management.
When Microsoft actively deals with this it's often through reorganizations. Clients are thrown together with clients in one reorg, regardless of the functionality. Then in a later reorg groups are made along functional lines, e.g. with Web technology in one area including both client and server. Each reorg causes huge disruptions and has high costs but eventually leads to better integration among the groups working closely together. The inherent problem of large organizations is that you can't have 40000 people working closely together. You must draw lines, only wherever you draw lines you are bound to create chasms.
In the email server group we heard reports of this problem from large companies who were customers. One customer quoted the figure of 70 feet -- the distance over which you couldn't work effectively with somebody else, the point in the graph where the efficiency most sharply degrades (although it continues to degrade as you increase distance, from one city to another, and again from one timezone to another). Of course our customers were looking to email technology as one way to address this problem but it was ironic that in developing this technological aid we were subject to the exact same problem.