I’ve had a case open with Microsoft support around load balancers in Azure, specifically around how the probes for the default probe work. There’s little information in the Azure MSDN documentation on exactly how the probes work. I did find this blog post on MSDN blogs by Kevin Williamson from 2013 that provides some details on exactly how they work, and confirmed with MS support that they do still function in this manner. In a nutshell, the default probe will perform probes against the Azure VM Agent on the server over HTTP and TCP to determine if the node is healthy or not, removing it from the LB if it does not get a successful TCP connection or HTTP 200 response. It’s disappointing that MS doesn’t provide more details on this in their Azure documentation, hopefully this is something they pull out of the blogs and add to their official MSDN docs.