Nope we didn't write any ChatProvider code.

 

I think the problem is the size of the application pool on our production server. We were able to load more requests on our other server before any operators were disconnected.