Raspberry Pi 4 with Mesa 7c81

It's totally strange. My thought is to try something to definitively say that it is not the 7c81 but I can not think of a way to do that. Open to suggestions for that or anything else!

Not a loose coupling

One more thing I just thought of...

I reversed the cutting direction of the ellipse. In the opposite direction the error still occurred in the first and third curves. So no matter which direction it is moving it botches the first and third curves

I still think its either mechanical or lost steps (drive related?) somehow.
Not really anything the 7C81 can do unless its fed a bogus trajectory that still obeys the
velocity and acceleration constraints and at the same time doesn't show up in the backplot.

I have checked everything mechanical. That was the easy part!

Also its too predictable in that it happens in the same places, curves one and three. I can manually command the machine back to zero and it is perfectly aligned. If I restart the program again it will do the same thing in the same place.

I just thought to try a G61 code...?

After looking at the .ini file again I did notice that the MAX_VELOCITY values for the x and y axis are different as follows

X
MAX_VELOCITY = 1.01666666667

Y
MAX_VELOCITY = 1.0

I don't think this could cause the problem...
The acceleration, ferror, and min_ferror are the same for all axis

I some was thinking mm instead of inches

Can you set ferror to .002 and min ferror to .0005
and try again?