Ask HN: Where do breakthroughs in computer science come from?

I know that plenty of breakthroughs have come from academia and research organizations, some have come from lone researchers who had a eureka moment, some have come from collaboration in the open source community, and some have come out of the corporate world.

More simply there's people who think about pure research topics, and people who are in the "real world" solving problems.

A typical scenario would be like when INRIA researchers developed the CRDT data type, and then it was applied by people building collaborative document editors later.

But it doesn't always work like that. Leslie Lamport had an academic background, but has said that it was his entry into the corporate world that gave him lots of real world problems to solve, which led to all his breakthroughs for distributed systems.

Linus Torvalds was purely focused on "real world" problems in the open source world and yet revolutionized version control systems as a consequence of leading distributed open source teams that required such tooling.

In a field like language design, it feels like there's a divide between academics who stereotypically have a bias towards things like pure functional programming and dependent type systems because of the formal and mathematical perfection, and then the open source world who care more about memory safety and data races because of the real-world impact on security, performance, and stability.

Can incremental innovation eventually build up to a profound breakthrough, or does it more usually require a bolt of lightning striking in a research team?

Do incremental evolutions in practical programming languages and algorithms right now constitute authentic breakthroughs?

John von Neumann in his essay "The Mathematician" said that there has always been a constant infusion of inspiration from the sciences and the real world back into mathematics.

Is it that intersection of the two where the most fruitful results come?