So, with a cartesian X/Y gantry backlash (as caused by a loose belt) on one axis results in circles with a flat on one side. Here I’m showing the desired motion (red circle) and actual motion (blue) of a 100mm circle with 5mm of backlash in the X axis.
The flat sides happen on the parts of the circle where the X axis motion is entirely lost due to the backlash.
With backlash on both the X and Y belts, you get 4 flat sides instead of 2.
OK, so what happens on a delta, like the Rostock? Here’s 5mm of backlash on the same circle, but applied to the delta tower centered above the image (The 3rd tower in the Kossel solution). The resulting “Flat” isn’t flat because the motion of the one axis doesn’t have a direct impact on the movement of the print head.
With backlash on all 3 towers we see a hexagon of flat spots.
Just like with the cartesian (not shown), the corners move around a little when the direction of the circle changes:
I hadn’t considered before that the print direction changing will show defects at different places due to backlash. This suggests a slicer could be modified to print perimeters in alternating directions to emphasize backlash issues.