I believe it comes from numerical methods, specifically for solving multidimentional differential equations. In order to solve a 2D differential equation, you need initial conditions at the boundaries; along the sides, you have one condition that must be met. At the corners, you have two conditions that must met, and frequently, that is where the problems happen.
Minor update: I should more properly have referred to those conditions as boundary conditions, that have to true all the time. In order to solve a 2d differential equation, you also need initial conditions describing the state of the surface initially (at t=0).
Project coordinator of the Generic Model Organism Database Project