In many cases developers having knowledge of the inner workings of the tools that they use is actually helpful. In other cases it's not helpful at all and they don't need to know those details. This presumption that game developers always know best is not a blanket rule. Developers building games in UE don't need to know every detail of how UE works just as developers of game engines don't need to know every detail of how the hardware or OS works. It's ok for some things to be black boxes.
I think you underestimate how important having knowledge is to a developer ...
One of the main reasons why Unity loses out to Unreal to PC/console developers is because Unity Technologies is far more
stingy than Epic Games is with access to their engines. Epic Games with Unreal could very well may have taken the same approach as Nvidia did with DLSS but they didn't and instead went with a community friendly approach much like the Godot project because they care more about the developer's experience/contribution/productivity than some streamlined procedures ...
There is no guarantee that a home grown solution is more sustainable than an off the shelf solution. That's the problem with this entire premise - the false belief that you can always do it better in house. Why would DLSS or XeSS be more of a technical debt concern than the tons of other 3rd party libraries that games already use?
Sometimes dependencies like hardware, OS, compilers, libraries, and engines simply pays for themselves because of a large community backing or contributions ...
It's not all about rolling out your own infrastructure and I can realize that but having the capacity to at least iterate, customize, or fix it is still empowering to the developers which shouldn't be discarded at all. There's tons of successful infrastructure projects that are very accessible to game developers like Clang/LLVM compiler (open source/all modern consoles have this available), Unreal engine (most popular PC/console game engine), Havok (most used commercial physics engine), Bink Video (acquired by Epic Games), Chromium (for displaying webpages inside app), and even artist content editing tools like Blender ...
So while industry maybe moving on from in-house expertise with infrastructure, DLSS doesn't have much in common with the above mentioned in terms of trends besides the fact that it's just another external dependency. The industry is dropping in-house solutions for more open and shared solutions because it's helps the industry retain important knowledge to pass on for
future entrants. It's an accomplishment by itself that the Unreal engine project has lasted for well over 2 decades and being open definitely didn't hurt it's longevity and in fact likely helped it become more successful when most software projects have development lifespans of less than 5 years. Can you absolutely say that DLSS will stand the test of time when it has nearly none of those qualities since it's author isn't interested in opening, sharing, or even educating it's users about the technology ?
If DLSS starts dropping in usage and it's only developer becomes an unwilling maintainer, I'm sure users would appreciate not having their projects being held hostage because some other party refused to update/fix their tools so you can obviously see why how DLSS has more technical debt than other 3rd party libraries ...
