FYI, have been in discussions with Microsoft on this today. MS engineer response to our Microsoft Premier Support case :

The behavior is considered by design and it is addressed in Critical Design Change Request.
We have scheduled a fix for Outlook 2010 in February 2014 cumulative update. A fix for Outlook 2013 will also follow, but Outlook 2007 will not be addressed because it is in Extended Support.

There are two possible workarounds for this behavior:

• First, and I wouldn’t recommend it, is to allow Direct traffic through the Firewall to Office 365.
The advantage of this approach is that it will be done centralized and it will impact everyone immediately.
Information about the O365 IP addresses Office 365 URLs and IP address ranges - Office 365 for enterprises

There are two downsides to this behavior:
o first you need to maintain the Firewall exceptions as IP from O365 server can change without notice
o second, winhttp traffic may go through both proxy and direct and this can cause unforeseen behaviors.

• The second approach is to use a proxy pac instead of manually setting in the proxy and the exceptions.
The advantage is that you will have centralized method to control how winhttp traffic goes and the issue will not be experienced anymore. However, you may have some challenges with deployment and connectivity topology so a single pac might not be fit. Yet you can configure proxy pac files for each site requirements and use an IIS server to store it.