If you are reading this the switch over to SQL Azure Virtual Path Provider worked (along with the fancy cache invalidation logic).  That means that the server-side code running inside of Windows Azure is actually coming from tables in SQL Azure.  Which means I can update the server-side code through a Silverlight editor without having to republish the WinAzure code.  Confused?  So am I.  But it is working   A more detailed post to come once the kinks are worked out.  I also wacked the Azure worker role which was costing me $80 per month so I’ll need to rework that into the web role at some point.