WT repair functions, such as the database size fix, are supported only on the leader. Requests sent to secondary nodes should therefore be rejected with a clear message.

The preferred approach is to follow the existing require_disagg pattern and perform the validation in __repair_config_set_command. This allows us to return the rejection message as early as possible, keeps the change on the WT, and provides a straightforward mechanism for adding the same restriction to future commands that must run only on the leader. No server-side changes should be necessary with this approach.