Under the shards' authoritative model, we have several copies of the catalog metadata across the cluster: CSRS, durably stored on the shard nodes (primary and secondary), and an in-memory cache on each shard node (CSR/DSR).

Under the authoritative shard project, the intention was to build a command that could make all copies consistent if needed. We didn't have the time and built a manual playbook instead. The playbook can be simplified by building this command.

I imagine this as a no-op "fix metadata" DDL that reads the metadata at the CSRS and then propagates the same across the cluster. It would also invalidate/rebuild the CSR/DSR from the local durable metadata on all nodes.

When an inconsistency is detected, the procedure will be to ensure/write the "correct" metadata on the CSRS, then use this command to propagate it.