which they [witnesses] have never denied [approving a hard fork]

This is not true. There have been at least two releases which were fully pushed out by the developers and blocked by the witnesses. There are other situations where developers have discussed the proposed implementation with witnesses prior to release and it was clear that some course correction needed to be made. This then avoided the situation of getting all the way to a final release which is then blocked (ultimately a waste of time and resources).

end up repeatedly failing to maintain the basic operation of that platform

Again you keep going back to "basic operation" when in fact basic operation isn't good enough. If all we wanted were that, we'd need almost no updates (occasionally there is a bug or security issue found but that has been very rare and usually addressed without too much disruption, and it would happen even less if we weren't pushing forward with enhancements). With no major updates or forks, basic operation would be damn close to 100% uptime and might well actually be that for years on end. But I personally don't think that is good enough. We need to push forward with improvements and that means taking some risks.

As far as taking responsibility, I felt that the post-mortem posts from Steemit after the HF20 screwups were pretty reasonable and that progress is indeed being made to improve the process over time. There wasn't even a public testnet at all until fairly recently, now there is. That's not everything, but it is something.

what I really want to see is the people who are responsible for them to feel some pain

Believe me, everyone involved feels plenty of pain from their investment being down 99%, and for some this is truly life-wrecking (whether or not anyone should get in that situation is another discussion, but nevertheless some absolutely are). If pain were enough, everything would be fixed by now but unfortunately it is not enough.