Both developers and testers are too clever for their own good. To them task that says “Just make everything work” is a challenge rather than an insurmountable obstacle.
Users on the other hand are usually sceptical. They are not the least bit inclined to solve puzzles when they’re just getting acquainted with a system. Sometimes, they resent it so much, that it turns into a full-on sabotage of your project’s implementation.
So where is that entry threshold? How does one estimate that scope of failures? What needs testing, if anything? Does that “little thing” really affect the product quality?
Let’s look at some examples and discuss practices.