In particular I've found race conditions and memory corruption to result in particularly fun "impossible" situations.
I try to say "I don't understand the mechanic by which that could occur, can you reproduce it?" and if they can then I have to figure out /how/ they can.
You should never deny the evidence. When you say "It's not possible", something in your understanding is obviously mistaken. Maybe your understanding of the evidence, maybe your understanding of the problem, but somewhere you're wrong. Your job now is to find out where you're wrong.
The correct response in such situations is "What am I wrong about?"
I've lost count of how many times I've seen the Can't Happen mindset delay resolution of an issue. It's a genuine problem.
> I've lost count of how many times I've seen the Can't Happen mindset delay resolution of an issue. It's a genuine problem.
> When you say "It's not possible", something in your understanding is obviously mistaken.
Not necessarily. Bear in mind that the "error" data itself can be wrong too, for many reasons -- some benign, some not so much. People can and do lie and make mistakes.
In the public sphere things are even more fraught. There are people who loathe $COMPANY and would love to see their services discredited. On the other hand, $COMPANY's legitimate success depends to some extent on people's perception of their reliability, so they have a right to defend themselves.
I think a reasonable response from $COMPANY in this case is "1) That's impossible", to reassure skittish customers, and "2) We'll work directly with the person having the problem and report back, stay tuned" to show respect and responsiveness (and potentially humility later).
If you were running your own company, paying the salaries of your employees and serving your investors, would you do otherwise?
It's loathsome that you ask.
Saying "We don't think this is possible on our end but are investigating to help wherever we can" is different than saying it is not possible.
I usually use "That shouldn't be possible" - whether it is possible or if it's user error then often depends on the maturity of the system.
On a new system pretty much anything is possible. On a system battle-tested for years by thousands of users the possibility of encountering program bugs drops dramatically.
This is where good supporters become very valuable. They will be able to learn the solutions to common problems that users face and determine if it´s user error, other errors like OS problems or if it's something new that should be investigated by the developers.
Of course if the bug is reproducible then it's a different matter. But any developer who doesn't take a well-described and reproducible bug report seriously should probable find a different job.
It's enough to get quite a few visitors who were aiming for the popular site.