Yes, making the corners 90 degree angles made them work nicely. Very cool zig-zag code.

Just checked it out. The top and bottom are leveled off, so it'll print without support. But the link connections with the sharpest angles are intersecting, so a few of the links on the plate probably won't separate properly because the geometry intersects.

For example, if the zig-zags have a 30 degree angle, that means that at the ends, where the belt takes a U turn as two 90 degree turns, they end up a 60 degree and a 120 degree angle, and the 120 degree angle turn is so sharp that it causes the links to intersect.

You can allow the links to avoid intersecting them by making them skinny (s=0.5, for example), but then the belt isn't (IMO) as nice. Or you can make the 'zig-zag' angle zero degrees, so the two corners are 90 degrees, which allows for nearly round links (s=0.9). But that's not as space efficient, and I really like your zig-zag space efficiency.

Previously I liked the look of fatter links, which can only flex 60 degrees or so )s=1.1), which is why the rows in my original are further apart. That wastes space between rows, so I'm sure that it's possible to be cleverer and more space efficient. Your zig-zag pattern really improves density, but it introduces sharp turns.

I've been trying to think of a decent algorithm for densely packing links (like your zig-zag) for link designs that only flex 60 degrees. I really like your zig-zag approach. I'm wondering whether with sufficiently clever math we could run the links around in a spiral-like pattern (perhaps in a rectangle, to fill a build plate)? That seems like it would eliminate all of the sharp turns, which lets the belts look as continuous as possible. But the math is pretty tricky. Perhaps after a good night's sleep? :-)