Many people know about the sunk cost fallacy. And they often think other people are stupid for getting sunk costs wrong.
But people often talk about real issues and call it a sunk cost when it's actually something real. So sunk cost claims shouldn't just be ignored as if they never matter.
A sunk cost is an investment in some kind of project which you already made and can't recover. It can include money, time, effort, etc.
Should you stick with projects you already invested in? Everything else being equal, no. If some other project is better than continuing this one, switch. The sunk cost should be ignored. You look at what continuing this project onward from this point is like compared to other projects.
Here's the real issue which people are sometimes speaking incorrectly about: We're presumably looking at a project with some large startup costs, barriers to entry, costs to finish the whole project, etc. If it was a cheap project, people wouldn't care about the sunk costs much. (Actually sometimes people eat food they hate because of the sunk cost of spending $10 on it already, even though they can easily afford food they do like. That's stupid.)
Projects are usually replaced by similar projects. So expensive projects are frequently being compared to other expensive projects. If I already invested $1000 in this project, maybe I'll have to invest $1000 in the alternative project, too.
So people complain about the "sunk cost" of $1000 already spent on this project. When what they really should say is they'd like to switch projects but the other project would cost $1000. If they'd change their thinking in that way, it'd be better. They're wrong. And they don't understand the sunk cost concept correctly (which already mentioned comparing continuing the current project from where you are to the new project from where it starts, which implies taking into account the startup costs of the alternative project).
But when you tell people to ignore sunk costs, you can be giving bad advice. They can think they are supposed to ignore the $1000 difference between the projects (both cost $1000 originally, but you already paid for one and not the other) because it's a sunk cost issue. Furthermore, this sunk cost issue didn't exist before they paid $1000 for the first project, since back then they had a spare $1000.