My takeaway is that the way the carabiner is loaded has a lot to do with the ultimate strength.

From an engineering point of view, I rather liked the gate latch design on the cheapo lookalike found at Wal-mart.  I wouldn’t use that one, but the overlapping notches looks like it would spread the load over more square whatevers than the design on the more expensive biners that had little projections on each side.  That kept being the failure point for the gate.

That overlapping design has the drawback that it doesn’t take any load until the carabiner deforms a little and there might be some loading range where the whole carabiner could stretch and take a set and not want to open anymore, but for me, that last might be a feature, not a bug.

My other thought related to the other end of the gate.  I think a couple extra grams making the gate longer so the pivot could have a little more material between the pin and the end of the gate would probably raise the point of failure.  It’s always going to fail *somewhere*,  but some of those failures looked like a slot.