Instead of trying one route after another, start async jobs all at once,
but each next one with an increased timeout.