A remark on the instability of the Bartnik-McKinnon solutions