[naniwaduni]

This might be a nice explanation of it didn't fail to agree empirically with reality, where in fact there is a significant cubic component which gives rise to the "something close to the 2.5th power" that GP suggests.