Exponentially distributed data is a great example of that. It's fun that the author picks 32-bit ints; I suspect that you just don't have enough dynamic range to really hobble interpolated search.