What about that using some post-processing techniques such as filters?
Does this cause any approximation issues with indirect lighting or is it completely fine to restrict ourselves to the area?
Why can't we try using things like low pass, high pass, and band width filters to help get rid of noise?
Is there an "optimal" number of samples to take?
How can we determine which directions have non-zero incoming radiance without trying all of them?
Wouldnt we have to sample all the light rays to see which ones have non-zero incoming radiance anyway?