When a drain request is received, all in-memory servers should be gracefully shut down. Additionally, after the drain has started, no secondary tenant server should be able to start. Jira issue: CRDB-21834 Epic CRDB-23559