it does not mean unbiased. you can do brute force in every renderer in terms of meshes, lights, shaders and textures without optimize things that would take alot of time. instead of this you have a big renderfarm and put the shot in and let it render 2 weeks. this is possible today because of the low hardware prices and the high power you get for it. when you look how big todays renderfarms are and that rendertime is not a factor anymore, then you know why they go this way. we reached a point for which less limitation making things impossible. particle simulation is one thing which would need much more counts then currently possible, but lights and geometry wise we can do alot of thing now. with a big farm you can do what you want without to cut down your shot in several layers to make it to render because it does not fit in ram, as an example.Eugen wrote:'not thinking about such things' would mean going unbiased, you mean that? That's just too slow mostly, even on renderfarm. Although some render animations unbiased, it's uncommon.
Would you want to spend a small fortune on a farm, or save time/money in the end by going through optimization?
That's why Redshift is a biased renderer. It's targeted at small shops and freelancers as I understand, but of course there's no reason why it should not be usable for big productions.
and for the unbiased thing in general, its hard to define such term today because also the unbiased engines using alot of optimizations to converge faster. they filter out values which will not have a big influence in the final solution or filter them out because it would take alot of time to solve them. or they needs some tricks and hacks because its not possible to solve the problem in a different way today. with this definition in mind, it would be possible to code a unbiased production renderer, but i think it would make no sense because you would define a "end" of rendering from start to make it "fast".