InstanceNotFound error after nova delete despite instance appearing in server list

After issuing a nova delete command, the instance remains visible in openstack server list output for a period governed by the oslo.cache expiration time. Subsequent attempts to show the instance with openstack server show return an InstanceNotFound error, creating a mismatch between the cached list view and the actual state.

The goal is to determine whether the observed stale‑cache behavior is an expected trade‑off for reduced database load or a configuration gap that should be addressed by tighter cache invalidation on delete operations.

What is the recommended cache expiration setting that balances consistency and performance? Should operators rely on manual cache flushes, or is automatic invalidation feasible without impacting Nova API latency?