it works. notes: i prototyped two versions simultaneously: a 1.6 mm thick simplified design (shown in the photo), and a 3.2 mm version (originally shown in the CAD models) 1.6 feels like the best choice -- a balance of weight/cost vs. local strength. i don't think global strength changes much between the two thicknesses i knew torsional rigidity would be a challenge with the flexure-snap design, and it does lack here. this is because the required clearance to ensure the flexure-snap fit works allows the system to rotate torsionally and parallelogram additionally, on this torsional rigidity: this parallelogramming would happen anyways even if the tolerances were nailed perfectly. this is a byproduct of "portal frames", i.e. frames that do not have bracing to resolve torsional/moment loads into axial forces acting on chords torsional rigidity solutions: a. create more moment resistance at the snap fits (i don't wanna, complicated/expensive) b. add "webs" -- i have some crazy ideas for this c. trick the snap fits into acting in a way to do A (frictional fit). we can figure out in 15 minutes of FEA if this will work prior to getting parts released d. something crazy like long additional flexures that act like webs when they're pulled into place in opposing chords (unlikely to work if they ever need to act in compression due to buckling) the center holes should be slots, but i wanted to prove the concept before going deeper into detail design cost: i think it's less than equivalent 8020. i need to do a better study to prove it. but things can only get better from here with refinement fun note: the snap fits were the main axial loading failure mode (buckling failure), but by looping them back on themselves i'm able to put them into tension, removing this failure type finally: i think i'm keeping the flexure-snap fit locating design. it was a pain for install because I had too many of them, but they can work if i reduce the count. i will try reducing by 1/3