You usually don’t actually need the best from an arbitrary sample, you need better than $threshold.
Aiming for best from sample when you have $threshold already could just result in opportunity cost.
Be it in hiring, dating, researching holiday destinations or anything.
The form of the solution also suggests a "good-enough" variant, useful when speed or simplicity is more important than optimality: reject the first option you find that genuinely meets all your needs. Accept the next thing you find that beats that first option. It is kind of surprising how often having this sort of approach in your back pocket can come in handy.
1. The problem only deals with optimising finding the best candidate, the "optimal" solution therefore has a high probability of picking a really bad candidate. In reality this failure mode is quite unacceptable.
2. While many real problems will have some opportunity loss by waiting, it is quite extraordinary that options are presented in such a linear fashion. In a real hiring situation you'd generally be able to go back to a candidate you interviewed earlier and hire them.
Choosing one which you know not to be the best is not considered any better than choosing none.
This may seem somewhat contrived, but I tried to come up with some tech analogy.
Let's say that you have some analyzer which takes in a finite amount of physical matter, and then produces some dataset. The analysis is a destructive process, so that the analyzer will only be able to produce N different datasets - until it runs out of the physical matter it is analyzing.
The datasets are huge, and will take up all available memory. Before the next dataset is finished, you need to decide whether to store or delete/discard the current. The permanent storage can only hold one of these datasets, and it is impossible to write over it. Once you have stored a dataset permanently, the next ones will get discarded as soon as they've been generated.
Or something like that.
Edit: Wouldn't surprise me if there's some embedded system / machine out there that's bound to select stuff like that.
What's "really bad" here and what is the probability of picking a really bad candidate?
If you somehow still have to immediately hire or reject all candidates and care about expected average, instead of just finding the best candidate. The optimal solution will have you reject fewer candidates outright, and then have the criteria for how good a candidate should rank to get hired lower gradually throughout interviewing, and drop sharply towards the end of the candidate list.
So I guess that's a use.
I believe there's a chapter in Freakonomics that deals with this problem and, if I'm not mistaken, talks about a situation where you might want to figure out some of the best places to eat in town without needing to visit all the places in town.