So wouldn't the correct course of action be, in this case, to close this bug report manually?

If not, then just tell me. I'm only learning this stuff.

I think we have now established that 1) he is not going to reproduce it, ever. And 2) the information provided seems to point to a problem with crypto, not with ext4, not with Sigil and not with his hardware, but 3) we can never be sure and this isn't going anywhere ever.

That is, if we think that what broke, broke essentially (the problem, whatever it was, broke everything that it could, namely all encrypted filesystems) and not accidentally (it could have broken any filesystems, but it just happened to break these particular ones this time).

We have also established that 4) there is no way for Devuan or any distro of any size to reproduce a problem like this, also based on the information given, but also because no distribution is going to start a hunt for a random problem based on one report. Even if there was widespread property damage or loss of life, there would simply be not enough information to do anything. There simply isn't anything obviously wrong with ext4, and it would be his problem to prove otherwise.

I personally feel that a distro does not look bad if it can't do anything about this. But a distro might look bad to some people in some cases if it has many critical bug reports hanging endlessly, because someone might think that there are serious problems the distro thinks it ought to fix but for some reason hasn't been able to. It muddies the waters, so to speak.

Do you have some kind of emotional attachment to AI or Simulation Theory? If not, then why do you imagine and insinuate that those who have different ideas are motivated by strong emotions? Such as derision. There is simply, plainly, obviously no reason to believe that any AI running in today's computers could do anything but remain artificial.

Did this happen again or did someone find the cause? I mean, the bug report still exists. But what are the chances of it ever going anywhere?

I am not especially interested in this particular bug, but I have always wondered what policies distros have for hanging bug reports. Can a report be in a moreinfo state for years? Has everyone forgotten this, or is someone trying to reproduce it?

I take it that in reality there wasn't an attempt to reproduce it, neither with a tainted nor with an untainted kernel. I am sorry for the loss of data that happened, but this stands now, four years later, as an act of God, strike of a cosmic ray or a butterfly flapping its wings in Japan and causing a data loss halfway across the globe.

Bug reports by definition can't do anything for disasters that probably only ever happen once. Unless the cause is known beforehand.

If there is a known cause, I'd suggest that something happened to the crypto key. It was altered in memory somehow and some important part of the disk's data or the on-disk representation of the key got overwritten. That wouldn't explain though why two drives were lost.

Were the disks physically fine afterwards?

What really happens is one thing. How it is marketed is another.

I am surprised that so many people are simply technically illiterate when it comes to how an AI works. That, or I have studied psychology and odd metaphysics far more than the average person.

Intelligence or intellect literally means the ability to make sense of something in relation to some purpose. All animals have what is called passive intellect that picks out relevant things from sensory input, helped by memory and instinct. A computer does not have instincts but it does have memory, and this is what an AI basically emulates: an animal without instincts. No AI, just like no animal, is aware of any purpose, but mechanically matches its input to the contents of its memory (and instincts, if any).

Humans have what is called "possible intellect". No animal and no AI has ever had any idea what should be or what could be. Therefore an AI can never literally imitate anything, any more than a line on a paper "imitates" the path taken by a pen. No AI has any concept of "truth" or "lie", any more than Pavlov's dog "lied" when it drooled after hearing the bell, making an innocent bystander think that maybe it sees food somewhere.

The thing that an AI does impressively well is that it smuggles Aristotelian logic into what otherwise would be a meaningless jumble of words and probabilities. Not because it understands logic, because it doesn't, but because the input material is sometimes ordered accordingly.

When people used to crack passwords hashed with md5, I found it impressive that someone could throw entire dictionaries and libraries at them. If your password was any phrase from Shakespeare, Tolkien or Star Wars, even slightly modified, if the hacker came up with the same rule you did, he could reverse it almost instantly.

In these language models today there is so much capacity that the English language simply is no match for it. There is no need to feed the AI all literature ever written, because after a few thousand books there won't be that many new patterns or combinations. The sheer capacity makes it possible to emulate the "understanding" of such concepts as truth or falsehood by simply designating certain word combinations as one or the other. An AI could describe "pike is a fish" as a true statement, but you could as easily ask it if "pike is a fish" is a left or a right, up or down, fast or slow statement and get an idea of the logic involved. It could for example say that it's more of an "up" than a "down" statement. Because it operates purely by numerically weighed assosications that simply give the appearance of logic or understanding.