I'm reading Horsman et al. "Surface code quantum computing by lattice surgery" and I'm wondering about the rotated surface code.
Consider Figure 13:
This is supposed to have distance 5. But in (c), an $X$ error on the top left qubit and the qubit beneath it would be undetectable (brown plaquettes are $X$-stabilizer measurements). If there were a stabilizer measurement on every outside edge, this problem wouldn't occur. What am I misunderstanding?