__disagg_step_up does not currently acquire a schema lock, but perhaps it should to ensure that __layered_create_missing_stable_tables does not conflict with any concurrent schema operations.

It might not be necessary to coordinate with creating new layered tables, because the node is at that point already a leader, so the stable component would be created with it. However, there is a possibility that it might conflict with dropping a table.

If the answer is that we do not need to acquire a schema lock at that point, it should be documented explicitly.