Before waiting for range deletions to drain, the CheckOrphansAreDeleted helper waits for migration coordinators to drain.

Under some conditions, migrations that hit errors simply clear the filtering metadata and are expected to be lazily recovered. Recovery can happen either on step-up or upon serving the next query targeting the namespace (that triggers a migration recovery as part of the metadata refresh).

The CheckOrphansAreDeleted hook must take into account that migrations may not be recovered if there are no queries or step-down+up. A possible fix would be for it to proactively try querying a namespace that has outstanding migration coordinator documents.