Talos is great until it's not. We ran into Ceph IO speed bottlenecks and found it was impossible to debug ("talosctl cgroups —preset=io" is a mess) because the devs didn't want to add an SSH escape hatch into their black box OS. Our Talos nodes would also randomly become unhealthy and you have no way of knowing why. Switched to PXE booted Alpine linux with vanille k8s, and we had a much more stable experience with no surprises, and the ability to SSH whenever we want has been hugely helpful.