I made a post about Facade occlusion and culling on the forums here. The occlusion is much more involved, so for now I'm leaving it alone and talking about the seams on the edges of faces.

private static final float zFightOffset = 1F / 4096F;

There is a line here that causes the seams, but I was wondering if there is a reason for its existence, because changing it to

private static final float zFightOffset = 0F;

fixes the problem. I know not everyone has the same hw as I do, and there is a possibility that the 1/4096 padding might be there for a reason.

It's hilarious looking at the log for this. I get it's an annoying thing to write, consisting of lots of math and effort, so I don't blame you or anyone else.

Fixed in 8.0.x.

This can either be fixed by using a super-complicated set of multiple passes with a "model bisector", or with a custom offset on a per-side basis. I wonder which is easier... :P

Wouldn't it be possible to do partial facade edge occlusion, though?

a) marking facades which are opaque vs. ones which are not with a helper function,
b) culling edges between facades which are opaque or have the same block/meta, as well as between opaque facades and blocks with isSideSolid() set right?

duplicate, can be closed

This is the zFightOffset - sometimes, in very rare cases, adjacent cuboids start "fighting" for priority and glitch out on the edges, which is quite ugly. This would probably be fixed with culling non-rendered sides, and is something to think about.