By plastic tabs, do you meant the “ears” of the MX-64, where the M2.5 bolts go into the brackets?
Yes, that is a weakness. In practice, this robot did very well along that axis in 2019, so I’m not worried, though! (I worried about this when I first made this nut-less design.)
This is, as far as I can tell, a weakness of the MX-64 servos, because even with brackets that bolt onto the full flange, the weak point under torsion is the nut on the underside. In practice, I think there will be little torsion, though, because there is boxing between “upper” and “lower” layers. (There’s no diagonal bracing between the layers, though, so there is that. But the yield strength of the thinnest member is > 900 pounds, so not too worried about that.)
The XM/XH series seems to have made this better by using inserts and larger contact area.
Btw: everything made by me on this robot is cut out of 6061 aluminum, and bolted with 18-8 or black oxide M2.5 or M3 bolts. One of the main goals is to use zero nuts, and instead put threads in all the right holes, and apply thread locker as needed.
What caused trouble in 2019 was walking on concrete. The concrete + steel + metal frame was stiff enough that something broke in the gearbox of one of the servos, and its encoder sensor came loose, which in turn caused the robot to bend its leg inwards, which in turn caused the (expensive Maxxon!) motor to damage itself through overheating.
Solution: Only walk on the PVC foam flooring (Or add rubber balls instead of steel balls at the toes.)