Because of the intermediate image updates. Showing the intermediate image, and keeping something ready to show you at any moment, requires extra resources and CPU time. F.e. color mapping and gamma correction need to be performed for every update; this is a small thing but across many pixels it adds up. The adaptive sampler needs to do this just once after a bucket is complete; the progressive one needs to do it for every pass through the image (there were about 1600 of them for this render).
The difference, if projected to a complex scene with longer render time, is quite big.This is not related to scene complexity at all. Only image-related properties have an effect on this (i.e. resolution, number of render elements and number of passes through the image). So for more complicated scenes, this additional time will be actually still the same and will be a smaller portion of the total render time.
Is it just about the progressive is relative new and still being optimized?There are some optimizations that can be done for sure, but there is just a bit more work to do generally.
Best regards,
Vlado