Besides that, many critical third-party resources (operating systems, niche professional libraries, etc) are still designed for C or C++ integration and often written in those languages themselves. When you use C++ to integrate them into your own project, you don't introduce an extra layer of bugs, limitations, or performance issues associated with bridging and marshalling and can more easily trace and debug issues that arise during or after development. Further documentation and support often presume you're using C or C++ and so you get the best insight into how they're meant to be integrated and the most help when you run into trouble.
If you're not already comfortable with C++ or not working in one of those scenarios, it's less warranted, but the two factors above account for plenty of real-world projects.
I’m mixed about this statement. Yeah, I had to write some pimpls to make cross platform API for multiple OSes.
And I’d also say modern C++ got nice compile time checks and makes safer code. Still, it doesn’t reduce potential of bugs or security issues due to race conditions, or direct memory interactions in comparison to writing in Swift or Rust as an example and binding what you need.