“…Consider the D-walk π 1 = π 1 0 π 1 1 , where π 1 0 ∈ ∆ B and π 1 1 is a shortest path of row-edges from (x 0 , 1) to (x 0 , y 0 ), in comparison with the D-walk defined as the shortest Type A D-walk from (0, 0) to (x 0 , c − 1) (built as in Lemma 7 and Lemma 9) extended with the additional edges ((x 0 , c − 1), (x 0 , 0)) and ((x 0 , 0), (x 0 , 1)), and then the path π 1 1 . The length of the former D-walk is r + x 0 + |π 1 1 | whereas the length of the latter is at most max{r − x 0 , 2µ − r + x 0 } + 2 + |π 1 1 |. Hence, as x 0 > 0, we can dispense with the D-walk π…”