Agreed. Having the isolation boundary at the VM/container layer feels much more predictable for experimentation-heavy workflows compared to relying purely on OS-level restrictions.
The local development loop part is interesting. A lot of developer tooling friction seems to come from iteration speed and workflow reliability rather than the core functionality itself.
A lot of the friction for us wasn’t Kubernetes itself, but everything around local workflows — reconnecting services, switching environments, and keeping forwards stable during testing sessions.