I've seen fantastic research sunk and forced through resubmission time and again because of reviewers that either don't understand the work (and why would they? they have zero incentive) or have pre-existing biases based on their own work (and are called in as expert reviewers, effectively functioning as gatekeepers in a subfield).
Bad deans/department chairs can make life suck pretty hard. This is usually not the case because they (in my experience) don't have that much power, but over some things like hiring they can definitely be as capricious as they want. The real killer are the people outside the department, on IRBs, grant review boards, etc.. In a good research-focused school they can be great; in worse schools they can be tinpot despots.
Reproducibility in particular is insane. In computer science virtually no systems are released at the time of a paper submission, so papers describe things that may or may not exist. I'd say this is less due to untrained/inexperienced grad students and more due to a combination of constantly rushing to publish, unclear institutional regulations regarding releasing artifacts such as source, and inability to focus on cleaning things up for a release.
Academia has serious problems at its core, and in a lot of places, it's obvious that startups and industry is outpacing it. I think this comes down to an incentive problem -- when there's no incentive to build things that work, you tend to build broken things.
Let's be realistic here. How many startups actually do any sort of research at all? By my estimates, the number is very close to zero.
How many places in industry? You can cite the big labs, but the majority of science is still done by universities. While academia has serious problems, lets not fool ourselves into thinking that industry is any kind of substitute.
Does git count as important research? Does CoffeeScript?
Apple does too, but unfortunately they lean towards keeping everything a trade secret rather than publishing (excluding their patents I suppose)
It is a race with very strange incentives this is the world where a Nature paper makes your career but publishing nonsense has no negative effect (in fact there is a positive one, you now have more publications padding your resume). You have to go out of your way like manufacture data and get caught to experience any negative consequence. Doing sloppy work is the norm.
I wish it weren't true and it could be that the problem is my perception.
As for the comparison with the industry: you can always go and try to work for a different company that does things your way, but guess what, science is all the same. You'll get the same reviewers, same grant application process, same publication process, same ranking methods no matter where you work etc. You are stuck with the system and you can't reform it.
Unfortunately, they are the norm. There are places when this doesn't happen, or where you have the privilege of working with an extraordinary professor who manages to shield you from this.
That's rare, though. #3 isn't constant and #2 is not an entirely accurate description (or at least not that widespread), but by and large, the portrait is very accurate.
I miss the intellectual challenge of academia. I'm in a pretty high-end job and it's nothing compared to what I used to do. But the working conditions and the colleagues? Not in a million years. I'd seriously consider selling hamburgers than returning there, and I have worked on minimum wage before, I'm not talking out of my ass on that.
Isn't culture fit great? I mean, if you're a social psychologist you won't have that problem, they have less than 1% Republicans.
I'm not arguing that the things you describe never exist in academia, but they're the worst examples of when things go wrong. They're not the norm.