1: \begin{proof}

2: %%   We first pick a shortest path $\gamma$ in $C$ between a vertex in $C_1$ and its counterpart in $C_2$. Now, for every vertex $v$ in $V$, there exists a unique geodesic in $C$ connecting $f(v)$ and $g(v)$ without crossing $\gamma$. Since the shortest path $\gamma$ winds at most a half turn around the cylinder, this geodesic winds at most one and a half times around the cylinder, see Figure..., and therefore can be cut into three geodesics doing less than a half-turn on the cylinder. These are shortest paths in $C$: for an intuition, if $\varepsilon$ is small enough the half-collar is very close metrically to the cylinder. This can be checked rigorously using the explicit metric in the universal cover of the cylinder.

3: 

4: %% There remains to prove that these are also shortest paths in $P$. By

5: %% choosing $\varepsilon$ small enough, our shortest paths are

6: %% arbitrarily close to subpaths of the boundary cycles doing at most a

7: %% half-turn. Therefore it suffices to prove that these are shortest

8: %% paths in $P$. This results from our choice of metric $m_P$: the length

9: %% of the shortest path not homotopic to a boundary connecting two points

10: %% on the same boundary of $P$ is lower bounded by the lengths of the two

11: %% ortho-geodesics cutting the equilateral hyperbolic hexagons into two

12: %% pentagons. Applying formulas for the lengths of right-angled

13: %% pentagons~\cite[Formula~2.6.17]{t-tdgt-97}, we see that this lower

14: %% bound is bigger than the half-length of a boundary cycle.\note{to

15: %%   explain more, and needs a figure}

16: %% \end{proof}

17: