Currently, startup_recovery_for_restore_restarts.js only waits, after inserting into a sentinel collection, for a log message that indicates an unstable checkpoint has completed. However, we must also wait for the log message that indicates the start of checkpoint, to make sure the insert is covered.