On Google's cloud, at least, VMs don't get killed for planned hypervisor downtime, which is a potential differentiator for them. Of course, this should matter a

On Google's cloud, at least, VMs don't get killed for planned hypervisor downtime, which is a potential differentiator for them. Of course, this should matter a lot less for workloads that have been explicitly designed run on Kubernetes...