Back to Back booking optimisation

I have a hunch that a number of predicted good sets do not become reality due to extensions and early pickups. We should investigate how detrimental this is to optimisation. And if found to be having a significant negative effect we should investigate the cost of applying a "buffer" to the end of bookings. A 30 minute buffer for example might be worth implementing on the assumption that bookings will return late. When the user returns, the buffer dissapears and the system shuffles all the bookings to refind the new optimimum.