Disk failures don't neatly wait for idle times so investing in a special vdev (which can be added to an existing pool as the article points out) should be well worth it.