This issue occurs due to a High Availability event. High Availability does not consult DRS rules for affinity or anti-affinity when a virtual machine is powering on. A DRS can migrate virtual machines to satisfy affinity rules after the virtual machine is powered on. However, this is not an option when using Microsoft Clustering, or other virtual machines using RDMs. This may result in multiple virtual machines with anti-affinity rules stuck on the same ESX/ESXi host that requires a reboot to migrate.