It sounds like a lot of this just comes down to the fact that our processes used for producing cylindrical cans aren't optimized for cuboids, and I won't argue with that.
However, part of me thinks that given some time and ingenuity, we'd come up with techniques that are better suited for the efficient production of cuboids. In other words, most of the things you mention are conceivably solvable by the right tooling (i.e. fixed costs). If that were the case, I'd have a hard time believing that marginal/per-unit costs would be significantly higher, and I think it would be interesting to look at the savings in shelf and transportation space compared to any of those increased production costs.
Cuboids can surely be done, he isn't arguing that. He's just observing all the ways in which cylinders are a very "elegant" solution, with a natural fit to metalworking.
I think metalworking has to be the key point of efficiency, because anything that comes packaged in cardboard (cereal, crackers, cake mix, powdered detergent, etc.) is a rectangle.
I would guess that cardboard is less expensive to source and work, so they can afford to take the extra effort to make it a box.
Cardboard's probably easier to form into rectangular shapes than cylinders - if you look at the way cardboard boxes are constructed, the joins are all done by overlapping which doesn't work so well on curved edges.
However, part of me thinks that given some time and ingenuity, we'd come up with techniques that are better suited for the efficient production of cuboids. In other words, most of the things you mention are conceivably solvable by the right tooling (i.e. fixed costs). If that were the case, I'd have a hard time believing that marginal/per-unit costs would be significantly higher, and I think it would be interesting to look at the savings in shelf and transportation space compared to any of those increased production costs.