You would expect this to be true if returns to scale are concave after some point. Intuitively this makes sense, for anything, after a certain point more of it becomes bad. If production returns to capital are sub-linear at very large (macroeconomic) scales then it would be surprising to see g >= r for any long period of time.