API issue: Actor.node_vms_with_local_snapshots is questionable

Is it allowed for snapshots to be propagated in the background? I have to believe yes. So if I have a 5K cluster with thousands of actors, is every actor going to be updated to list every node? I can't imagine that is the intent.

Search for `Actor.node_vms_with_local_snapshots` in the Go code and inspect all call sites to see how snapshot lists are propagated. Then trace related background propagation/scheduler logic to verify whether actor updates are intended to cover every node for every actor. Locate existing tests around actor or API snapshot responses, and finish when expected large-cluster behavior is covered (and updated if semantics need clarification).

Written by the indexing model from the issue text.