Oh, I agree, but that’s the crappiest thing about software: you don’t always prioritize the things like that. Most of the times it’s just cold, hard business sense.
It’s kinda like that scene in Fight Club where the narrator explains the “recall formula”, prioritization in software can be “heartless”, for lack of a better word. If the problem is hard, you might go for low-hanging fruit instead. If the problem affects only a small subset of your users, you’ll tend to prioritize something that affects a bigger share of them. If the problem is easy for your users to work around, you’ll tend to prioritize something that affects them more.
I’m not saying that AI getting stuck on a small rock or forgetting how to fight is any of those, although it certainly sounds like a hard problem. I’m just saying that prioritization in software development is not only complicated, but often leaves a bad taste in the devs’ mouth.
The game dev industry is even more complicated in that respect than most of the rest of the software industry. A typical game dev team has many roles besides programmers. There are so many different kinds of artists, it astounded me when I first got to work for a game studio. And therein lies the problem: even if you could afford to put all your devs on bugfixing (and if it was technically feasible, which it often isn’t), what do you do with the artists? Not only is it a waste of resources to have them idle, they’ll also get restless and maybe quit and go to a different studio. So what do you do? Either you have part of your team working on new content, or you reassign your artists onto a different project and risk not being able to get them back.
And there’s always business, too. Imagine a hypothetical product idea for software in cars that detects that the driver is not paying attention at the same time as it detects that there’s a red light or a stop sign coming up, and sounds a loud beep to alert the driver. Now imagine a hypothetical situation in which you have to demo that technology to see if the car manufacturers are interested, and your bosses decide to change how the technology works: instead of alerting an inattentive driver about a dangerous situation, it now tracks whether the driver is looking at a billboard with a fast food chain logo and pops up a recommendation on your dashboard for the closest fast food joint in that chain. I hope you understand just how bad a hypothetical dev working on that would feel in that hypothetical situation.
At any rate, I apologize for derailing this thread and pouring my heart out here, but sometimes it gets kinda personal for me and I try to raise awareness of how things aren’t always the way the devs would like them to be, either.
Honestly, thank you for that. I can get really abrasive sometimes and it’s always nice to hear that I manage to avoid that occasionally 
