Hi Grendel, also some additional notes:

You'd probably get some better quality by bumping up "Canvas megapixel resolution" to something like 30 or 35 at export time, that will make a finer grained depth map to be used for determining visibility.

But since that will consume additional memory itself, you'd probably need to both set the display mesh density to very coarse and also prepare a lighter file with the back half stuff deleted as well in order to do that.

You also might get some improvement by scaling the whole thing down by 1/10 or so in size, right now some of the bias tolerances to prevent "self shadowing" type artifacts are just hard coded and not adaptive to the overall base model size like they should probably be. So keeping things more like 30 or 40 units across or so as the overall size may help to avoid some more drop-out type areas also at the moment.

- Michael