a.moore, Benonymous, and kikiturbo - I brought this concern up to my cage guy after it was pointed out and he stated that the vertical bar is support for the horizontal and the angle pieces are there to help hold it square. Only experience I can go off of is when we built trusses in 7th grade and the design of the truss would determine the load the roof could take. I know that he is not doing this out of laziness or difficulty of a weld.
I didn't want to push the issue without running some numbers in case the stiffness loss was negligible. All elements have been set to beam (ends can carry bending load) and the material is 1" x 0.065" round steel. A 1000 lbs load was applied to node 4, fixed node 1 in the X,Y,Z, and node 3 in the X,Z.
The first test was with no diagonal and it is bad:
1x1x0.065 24inx24in square no diagonal.JPG [ 82.91 KiB | Viewed 952 times ]
Node 4 moved about 1.9" down.
Next the diagonal was added but it was located 3" away from the nodes:
1x1x0.065 24inx24in square diagonal 3in from node.JPG [ 84.76 KiB | Viewed 952 times ]
Its much better than before since node 4 only moved 0.094".
Finally run the diagonal node to node:
1x1x0.065 24inx24in square diagonal to node.JPG [ 85.11 KiB | Viewed 952 times ]
Node 4 only moved 0.0165".
By running the diagonal the way he has, it is only about 1/6 as stiff as running it to the node for that load scenario but it is still about 20x stiffer than having nothing. It would take a lot more work to determine its effect on the overall frame stiffness but you get the idea that it does matter. With that said the book Locost frames are not the best from a chassis engineering standpoint and they do fine.