Interesting.
Since they are a fixed width, there is no error possible in setting up the width of the crossbars.
The L-brackets are simply locating brackets. At 3Nm, that is very low torque spec. It is not intended to hold weight, just hold the rack in place.
Pulling forces would be on the nut and the L-bracket and mounting tab. (3Nm is barely beyond finger tight with the wing nut)
At this torque spec, a torque wrench is critical. It should absolutely be included, and if not, have it installed by Tesla. That is probably the key factor in the install.
Breaking of the old brackets indicates there is a cantilever force on those brackets. The original bracket was welded with only vertical load in mind.
Full perimeter weld to stop bending forces. I wonder if the bracket itself is a thicker gauge steel?
I looked at the installation materials and can only think of several ways it could touch glass:
1. There is enough flex in the crossbars for the legs to bend inward, into the glass.
2. The load compresses the rubber feet, loosening the nut and giving enough play for the rack to shift over- thus putting weight on the glass.
3. Over torque the nut, bending the L-bracket(loosens the mount), forcing the foot into the glass. Put a load on the bars adds more flex to push into the glass.
4. Same as 3, except it allows side shifting to land one foot into the glass.
Since the feet are rubber, the rack will compress with load.
I think it would be prudent
1. Check the feet are not touching the glass on installation.
2. to torque to 3Nm, tug at bars in all directions per instructions.
3.have the rack sit in full sun (soften the feet),
4.put on your heaviest load, and then torque again to 3Nm.