This is the issue you’ll never get completely around with autonomous systems. A soldier can always figure something out, whether that is simply clearing their weapon, completely disassembling it to repair it, finding a new weapon on the battlefield or getting a buddies side-arm. An autonomous system will never be as versatile and capable of adapting to stuff breaking as a human soldier.
The major advantage with autonomous systems is that you can accept that they break and become dysfunctional in the field. You can always manufacture more, and none of your guys die when one of these fails.
With all that said, I would think you could get pretty far by just adding some arm that can slide back the bolt to clear/reload the weapon when you get a jam. Like 90+ % of the jams I experienced with the MG3 and HK416 were cleared by just doing that.
While I see what you’re getting at, I still like this XKCD. I work as a developer, and have also worked in more “handy” fields. The thing with planes, elevators, and basically all other physical things is that they’re limited by physics. A steel beam can’t suddenly decide to spontaneously fail or disappear.
With code, that can feel pretty different. With experience, I’ve basically learned to assume that there is always some edge-case I haven’t considered, that could trigger a bug. In a building, you can have redundant bolts, and over-dimensioned supports. A small mistake somewhere, a single missing bolt, won’t cause a catastrophic failure. With code, it’s different: A tiny, hard to notice mistake, can bring the whole think crashing down. Imagine if a plane could crash because the paint had a slightly non-uniform thickness…