Are we concerned about the fact that if we have two primitives at the same depth, then the order of evaluation now actually does matter?
Does blending from supersampling affect what happens when two primitives are at the same depth and order evaluation matters?
Is it because if two primitives have very similar/ same depths, floating point inaccuracies might cause inconsistent result with which primitive gets displayed?
How efficient is this method?
thinking about the implementation side of things, does this implementation assume we're only testing for depth via the z coordinate? if we're testing for depth along a vector, would we need to project the coordinates onto the vector space or something like that? or is it okay to only use the z coordinate?
When tie breaking, is there a common practice for which primitive gets rendered, or is it random?