We shouldn't roll out everything for everyone at once. Things will break. @ryscheng has been suggesting we use PostHog feature flags for quite some time, and I think this is the perfect use case!
How I see this working out is, keeping old resolvers live side by side to the new resolvers, and at the osograph router, check feature flags to see if the user is internal or not. If they're internal, serve the new resolvers (which need thorough testing), if not, old resolvers.
We can setup everything via the UI and, if needed, even filter Carl's user out in case he's doing some important work and some resolver fails in the critical path.