I used to think C++ was a very nice language. That changed when I tried to make a game.
Similarly, I used to think that C++ was a bloated mess, and that changed when I wrote my first serious application (which wasn't
In my own experience, high level language features are very alluring and enticing at first glance.
I don't think you can talk about "high level language features" as if all high level language features are a monolithic entity. The question, to my mind, is one of impedance match. The high-level features of C++ are far from perfect (which is one of the reasons why the standards committee keeps tinkering with them and adding stuff to ameliorate the problems caused by the last round of standardisation), but they are actually pretty good at what they do.
Unfortunately, what they don't do is what game devs want them to do. The features you want (e.g. reflection and reification) aren't there, and the features you have (e.g. inheritance) just aren't that useful.
I've worked in a lot of fields, but the two I want to highlight are visual effects and bioinformatics. I worked in visual effects some time ago. The one feature of C++ which helped enormously was namespaces (and by extension, classes-as-modules). Keeping namespaces clean is a "must have" on any medium-to-large-sized code base. Abstract base classes helped a bit, but we honestly could have done without them.
When I was working in bioinformatics, however, many of the high-level C++ features saved my bacon more than once. Yes, that includes templates; template metaprogramming was an excellent fit for the complex data modelling and advanced data structure work that we needed to do there. And when you aren't entirely sure what high-level problem you're solving (which is all too often the case when it comes to biology), being able to piece together reusable high-level components in different ways helps a lot.
The high-level features of C++ were developed for reasons. Like any system which evolved, there's a lot of cruft. However, all that stuff was added to solve real problems, just not necessarily the problems that you are trying to solve.