The Azure FAQ on-line specifically addresses planned (for maintenance) and unplanned restarts of VMs. Dealing with planned restarts should not be an issue. Unplanned restarts may occur when serious hardware problems are detected, which would probably create a data issue anyway. Azure automatically migrates the VM to a healthy host and restarts the VM in that case.
The recommendation from Microsoft is the following:
To provide redundancy, put two or more similarly configured VMs in the same availability set. This helps ensure at least one VM is available during planned or unplanned maintenance. Azure guarantees certain levels of VM availability for this configuration.
We have not tested any of this, but it appears to me that Microsoft has carefully considered what may occur and how to deal with these events. Sage has not provided failover within the MAS application, so the VM approach is really the only practical solution. Perhaps I do not understand all the technical issues involved, but I am interested in finding reasonable resolutions for Azure.