I don't think we disagree on much. Unity has failed to stick the landing on most features developed in the past decade. If you make small to medium sized games, and treat it like 2015-era Unity, it still works decently, but progress since then has been slow going. All the new-ish features since then, like Scriptable Render Pipelines, Networking and DOTS had a rough development path, with constant bugs, breakages and stillborn features, sometimes packages got deprecated before reaching production status.
As for the C++/OpenGL comparison, I could've just recommended any common sense approach using industry standard tools, like Unreal.
I'm just saying, in terms of code performance, DOTS promised a paradigm shift, this cool new way of structuring code and writing a dialect of C# promised heretofore unseen heights of performance. The reality was that Unity Burst code is barely faster than CoreCLR .NET code, and is slower than straightforward no-frills C++ (but at least you get Unity vendor lock-in). This performance matters when you're trying to code a core game system, like procedural shattering, CSG, or inverse kinematics, doing it in regular C# will bite you in the butt. I'd rather do C++ for the few things where performance matters (not to mention, I'm 99% sure that well-tested high performance open-source C++ implementations of the above problems exist), than let Unity take me for a ride with DOTS.
Just take a look at Unreal. I'm pretty sure most of the stuff Unity's still working on wasn't a problem circa UE4 release@2014, and most of their 2014 demos would work in a modern version of Unreal (not sure how much of a breaking change UE5 is, but definitely in the latter UE4 releases).