Further attacks on the mysterious TCP CLOSE_WAIT issue