People (including me) always talk about how memory safety without GC is the most important feature of Rust, but just as important is how easy Rust makes fine-grained parallelism. In Rust, you can often start off writing sequential code and then bolt parallelism on after the fact and have it all "just work" with improved performance, thanks to libraries like Crossbeam and Rayon and the way the ownership/borrowing rules enforce data race freedom. That's really powerful.
And if it doesn't "just work" then there will be a clear error message pointing to the problem instead of nasal demons haunting you. That way you can debug and fix the problem much more easily.