Got some more tests done last weekend.
I added 20lbs worth of steel weights to the sled. I can now get a good triceps workout every time I attach the sled 
The added weight has helped immensly.
I sacrificed a sheet of CDX to a rigorous tolerance test to see where distortion was the worst:
I numbered each of the squares and listed their tolerances below:
My conclusions:
I need to work on the accuracy at the top of the bed. The chain actually jumped on me during testing and I’ve been revisiting my risers a little. Apparently I had quite a pitch to my chains, and never noticed because I was never in the top left corner of the machine. The squares on this particular test were too high up and the sled also tilted (after it jumped a chain link). I decided to abort the top row and investigate the results I got on the bottom half.
I’m actually pretty okay with the results I saw across the bed. I still haven’t gotten the X-Y calibration right with the linkages, so I’m out somewhere about 1mm consistently, with the X being longer than the Y. On the far right and left squares at 1m out, I expected that I would start seeing distortion on X-Y because of my motor spacing. I can see some of that distortion creeping into the 750mm and even some of the 500mm squares, which really limits me to about 600mm (~12") off center for good accuracy. The Square tolerances reflect that distortion that is happening.
I have also been upgrading the sled:
I replaced the stack o’ washers with Derlin spacers, and also used 1" spacers to clear the nylocks above the now solitary riser panel.
So once I fix the sled toppling over the top of a sheet with a second fence, I should have 1200mm x 1200mm (~4’x4’) of reliable accuracy.
Still working on a video. Editing takes a lot of time.