Rhel 10 might come with fully immutable hostfs, by bifrost (We have no inputs from RHEL team yet on what exactly we will get in rhel10 in that regard)

That affects the way we configure and deploy containers, packages on the host, and how we roll out minor updates on EDPM hosts.

The crucial point is to avoid unnecessarily turbulence for running VM instances, as a compute host reboot requires live migrating all VMs.

With mutable OS in OSP, that used to be OK to fulfil this requirement with a manual action, because a reboot was *only* required during low level system changes, like kernel, or ovs.

With rhel10 immutability, we need to choose carefully in which way we provide updates to EDPM configuration changes. If we are going to do that for any EDPM container config bit, or a package change,  we must build an automation around orchestrated live migration of workloads, API cordoning, and rebooting compute hosts without users interaction, and without global cluster downtime, because that will become something that should happen on *each* minor update as of now.