After scale_up, restart the scheduler and wait for it to be ready before other services start. The scheduler's init containers run migrations, so this ensures tables exist before ingestion/extractor pods try to query them.