From what I remember in reading through the motorolaMotorola transistor data book a few months ago, switching transistors, as you stated, have a faster ft and because of this, they have a smaller linear region. Small signal transistors have a slower ft, but a larger linear region. I recently took a VLSI class that unfortunately only focused on MOSFETs. From this, I can only assume that the length of the N region in n PnPPNP or the length of the p region in an NpNNPN in a switching transistor is smaller so it's easier to make the depletion region large enough to make the transistor conduct. I would also assume the opposite is true for small signal transistors. Hope this helps