I don't think the value proposition of Rust over C#, Java, and Go is that it eliminates more errors. I think the advantage is it eliminates the same number of errors and performs better/has less overhead.
Lots of people argue that Rust improves correctness by way of sum types (including no implicit nil/null), affine types (addressing data races and so on), and (compared to Go) generics. I completely buy this argument.
Those data races are only addressed for in-memory accesses across threads.
If you are accessing data via some mechanism of IPC across processes, e.g. database access to same data rows without proper transactions, accessing shared files without locks, GPU memory data,.., affine types won't save you.
I agree with your post generally though.