The gist: render in 2 passes: fist, render fully opaque objects top-to-bottom so you can avoid drawing pixels twice in the same location, and then render the translucent objects bottom-to-top.
This should give big performance wins in theory (because you avoid drawing pixels which you don't have to) but sounds very complicated to implement. Also, how will this work with custom shaders?
Z-culling is explained here.
The gist: render in 2 passes: fist, render fully opaque objects top-to-bottom so you can avoid drawing pixels twice in the same location, and then render the translucent objects bottom-to-top. This should give big performance wins in theory (because you avoid drawing pixels which you don't have to) but sounds very complicated to implement. Also, how will this work with custom shaders?