As an approach I think this makes a lot of sense. First it gives the very best students an opportunity to really show their stuff. For the others, it gives them the opportunity to pick and choose the problems to show that they've at least mastered some of the material (generally, there is often more presented than can realistically be mastered of the course of a semester (between and within each course)).
(nested parens are (the best))
I had a history professor who gave a test in this manner: Below there are 3 columns of 14 items of interest, whether they be people or wars or treaties or whatever. You are allowed to cross out 2 columns, and then cross out 3 terms. Write what you know about the remaining terms. Your grade is out of 110 on a 100 point scale.
Now, I loved the test, because I excel at soft trivia tests... but did it really gauge what I understood about history?
Nope. It gauged my knowledge of history trivia for those terms. Understanding history is about understanding how these things connect together.
If I were a history professor, I would base the majority of students' grades on papers about the subjects we cover in class to discover whether they're actually thinking about history or if they're just trying to recite facts and figures.
In typical work scenarios you have more than 2 hours to tackle a problem, anyway.
This is computer science, and a data structures class to boot. Answering any of the questions required not only a deep understanding of the fundamental principles involved, but analytic skills (both in parsing the problem statement into a set of requirements and to identify problematic edge cases).
Any good class will teach more than they could possibly test on in two hours. So the alternative here is to keep students guessing about what will actually be on the exam? That's why, when I was a college instructor, I gave previous exams to my students (also: i thought not doing so would disadvantage less socially connected students; lots of old exams get passed around).
When I was studying, to me, the grade was just a benchmark to measure my progress against, but otherwise wasn't something that I was particularly concerned about (in second level, I only cared that I good good enough grades to get the Uni placement I wanted, but beyond that, it meant very little and means absolutely nothing to me now. And third level... I've literally never told anyone[2] what grades I got on my degree. It seems that having a relevant degree at all is a ticked checkbox, but that's about it. Once you have some experience, that's all anybody ever cared about, at least, for me).
[1] But I guess even if the spread/variance of grades was high (rather than everyone being clumped at the top because its too easy to get good grades), grades aren't a great indicator for many reasons, so the utility is probably limited either way.
[2] In the context of trying to get a job, at least.
2. A lot of the moral case for education debt is that it's good long-term value. People graduating from CS education without the education needed to get a good job don't get that value, making the moral case far murkier.