I can't really believe that you don't understand, so I would say simply test a Ci$co appliance or something like that. You'll clearly see in seconds that OpenWrt isn't a firewall, nothing close.
I just beg (moreso the OP), please don't keep trying to inquire or postulate if you don't understand clearly. I (and I'm sure others) don't want their routers having issues because someone got the notion it's a firewall and removes those rules from default.
Otherwise, this seems like something for upstream and various Kernel changes. But as I noted, most exist in sysctrl/Kernel already (i.e. ICMPV6 responses are Related traffic - unrelated to the firewall entries the OP highlighted). The OpenWrt is routing/forwarding that unrelated traffic in the firewall rule.
I also noted these are stateless recommendations in the RFC - but somehow the debate continues on stateful firewall rules? 
This is why I tried to highlight the sillyness of testing by making stateful firewall changes. Somehow, it was taken as some credence supporting your pont-of-view. The test would be to remove the rule, it should still work on Established traffic (hence, in code).
So then its noted (or rathar, my statements seem discounted by you) - because you see stateful rules?
I'm not sure how that's difficult or confusing (i.e. stateful vs. stateless), but I've tried my best to offer clarity. I do appreciate the level-headed collegiate discussion, but I think it's ran the course.
As I said, not sure how my statements were unclear; but nonetheless, carry on.
You missed my point (it almost seem on purpose and comical). I actually can't tell if you're j/k; but I don't beleive you'd persist if that were the case - so I wanted to respectfully formulate a coherent response to you.
Again, thanks for the scholarly discussion.
Perhaps I'm wrong, eh...it's possible too.