I don't want to drag this out too much, but I have been teaching at a university for a couple of years in a course that actually did involve handing in assignments digitally. I consider the people I was teaching (physics grads) technically skilled, or at least more skilled than average graduates, but I've witnessed a lot of struggling with the simple requirement of compiling the solutions into a single pdf. It's difficult for me to imagine that the average high schooler will be able to do this reliably. I might very well underestimate the "new generation" of students and they capabilities though, so I guess it's just a matter of having the right tools.
I can only speak to what I've experienced and seen; Whilst I'm a CS student now, I did politics as my main course in high school, and even before high school, handing in assignments as PDF was the norm. I think once it's normal it's just one of those every day things you hardly think about regardless of what tool you use. But I guess to properly make a judgement on it we'd need a much larger cross section of society to give their opinions, both students and teachers from many different fields, and after trying out both systems.
In the end I think we agree that the tool isn't the important part - at least unless it's in a field where learning how to use a tool is part of the goal. Question is just whether learning how to use the tool lessens or improves the main learning outcome and ease of working. In my experiences with all the things I've studied I feel it's made me faster and helped me organise my thoughts better than a pen and paper - but that's more anecdotal than anything.
Though it is a fascinating debate that could warrant further studying by those with power of the system, to see what is ultimately the best for students; perhaps a hybrid solution really would be optimal, I don't really know.
Regarding the CAS topic: Yes, I think generally, computer science simply isn't as demanding regarding calculus, as physics or pure math (And I don't mean to sound superior about it, I am very aware that when it comes to discrete maths, CS grads are way ahead of me). I think it's great that it worked out for you, but in some fields you'll feel terribly lost without a solid calculus basis, as some of my friends back in my undergrad years admitted to me. I know it may feel silly to work stuff like integrals out manually (and don't get me wrong, I hardly ever do that either), but when you encounter, say, a derivation where integration by parts is used to simplify an integral, it gets pretty important that you have an intuition for these things. The intuition will never develop without seeing every step behind the scenes.
Hm. I think it depends which university you study CS at. I have friends who study CS at other universities, and by what I've heard from them, my university has much greater focus on math and theoretical math especially than theirs. But of course, regardless of where it is, CS is different from studying pure math at least, so of course not as mathematically demanding as that would be. But I'm currently in my second year and have taken Calculus, Linear Algebra, Probability theory and statistics, Formal logic and provability as well as computability, which is arguably more a CS than a math course, but I count it as mathematical since it was more math heavy than most of the purer CS courses.
And while we do a lot of stuff with CAS tools and the like, we're still heavily encouraged to du some things by "hand", still writing it out on a computer - often you also only get points if you show your steps - but you can still use a CAS tool to verify your result as well as use it as a writer - it's like a mathematical typesetting program that is what you see is what you get. Essentially you can use a CAS program like Maple as if it were Word for math.
I guess what I'm saying is that I agree that there's a lot of value in working some things out manually; I just don't think you need to do it on paper to get the desired effect, and I'd still say it's faster and easier to write out manually on a computer than with a pen, but again that's just anecdotal. I also think there's great value in doing it with a CAS tool though, so you can focus on the bigger picture and meaning of the results, rather than the often trivial steps of a calculation that can easily go wrong - You do Gram Schmidt process on just a slightly large input without making a fatigue mistake

. It's the same thing again, and again and again - it's not hard, it's just a lot of steps. For my linear algebra exam I did it by hand since we were supposed to show our workings out (written in Maple) but verified my result in Maple. I had made a mistake somewhere, but because of time constraints I just wrote that as a comment; I was sure in my methodology but had made some silly mistake somewhere, the correct answer was whatever it was, and moved on. Full points, cause they could see I clearly understood what I was doing since I still had all the steps but had just written a wrong value somewhere, clearly by mistake as I argued each step, and I recognised there was something wrong with the validation stage. I think this is a way of doing things that makes a lot of sense since you sort of get the benefit of both worlds, and it takes almost no extra time to write out your results and verify them since it's all within the same tool, and to me at least takes less time to write math in maple than with a pen.
No matter what though, in the end it's what works best for the individual, or in cases where it's impossible to make individual accommodations, the majority, that is most important and I can only speak to what works for me and the people around me whom I've spoken to about things like this. I'm very happy with Denmarks fairly digitalised education system though