Don't be quick to dismiss Powerset's indexing speed. It is relatively easy to improve the efficiency of a known algorithm. Likewise, it is easier to develop an algorithm if efficiency constraints are defered.
It is relatively easy to improve efficiency within one module by 10% and occasionally by a factor of four or more. If the average system performance improves by 10% with each change then a team could eventually improve efficiency by a factor of 2^10.
Yes, many small changes. Furthermore, this is a pessimistic case. If your average improvemented was 12% then you'd require significantly less improvements.
It is relatively easy to improve efficiency within one module by 10% and occasionally by a factor of four or more. If the average system performance improves by 10% with each change then a team could eventually improve efficiency by a factor of 2^10.