Parts of systems written in a language doesn't really mean anything for its adoption into mainstream. For example, Amazon uses Ruby heavily for a bunch of deployment stuff, but Ruby (sans Ruby on Rails that is in decline), is not really a mainstream language any more.

>Equating the existence of unsafe with C is laughable imho

Im not comparing it. The point is to demonstrate that unsafe exists for the sole reason of performance. In fast code you often want to directly access x[y] where x and y are variables, without having to run extra code around it. Its a well known computer science thing, as most of the code challenges given in interviews rely on this access pattern for optimal solutions.

And because of Rice theorem, a compiler cannot determine whether x[y] is always safe, as determining all the values y could take would involve running the program.

So as such, for all the advantages that Rust offers, you can have the same advantages with C with macros and LLVM extensions, albeit with less concise syntax.

https://www.microsoft.com/en-us/research/project/checked-c/

Similar arguments were used to justify Haskel about 6-7 years ago, and Haskel is pretty much dead in the water at this point.

The modern way to make a memory safe language is to focus on a high level language that doesn't require programmer to deal with memory directly, and then work on the compiler to make the resultant code optimal.