No, I assume that the nodes are only located in a topology that is connected according to simple rules, while the Euclidean space with its angles emerges from the graph at a large scale. Rather than grids coming together from random directions, Imagine instead first one grid and then the second grid grows over it according to a simple rule with a default inclination. So one grid is the parent of the other grid and then you keep on adding further grids until you get a tree of grids all pointing in many different directions which leads to the rotational symmetry of euclidean space if you measure graph distances on the large scale. The drawing is just the simplest example where alpha is not irrational wrt pi, but if you change the x:y ratio from 1:2 to some other ratio e.g. 2:3, you'll get an irrational angle wrt to pi.