On multiple returns in the random-walk problem