I had a similar problem. We are deploying using a scripting environment in Azure, and it turned out that the "old" WebJobs (running code with the old version of the entity model), which is still running in the queue. These are jobs where they then retrieve messages and consume them. To add to the problem, an exception was used in an attempt to catch, and WebJob status was successful.
Check if you have a scene environment (add -stage for the name Webapp), and if so, log in to the Azure Management Portal and stop them.
Note that stopping Webapp is not enough; you must stop WebJobs immediately. This is done (on the new portal) under "Settings-> Web Applications", and then right-click on the name of the website, selecting "Stop".

source share