A constrained move keeps only the snap correction along its direction, and that correction exists only when an edge ends within a quarter cell of a gridline. When no edge on those axes is that close, the move does not snap.