Service Healing is an important Azure characteristic.
Microsoft Azure constantly checks the health of the nodes and virtual machines (instances). Both the software and the hardware of nodes are monitored by the Fabric Controller. If a problem occurs, for instance, a cooling fan malfunction, the host agent will inform the Fabric Controller about this. The Fabric Controller will then move virtual machine from the problematic node to other nodes. As Microsoft Azure does not have the ability of Live Migration like in Hyper-V, the virtual machines will be shut down, reallocated to a different node, and then restarted. The fault node is marked "out for repair" so that Microsoft staff can replace the node. This process is called service healing.
When a virtual machine is restarted, the IP configuration will be reset. As you learned earlier, it is not wise to manually set the IP address of virtual machines. This will work until the virtual machine is reallocated to a different node...