I agree with you overall but there's a huge caveat: you need to make sure whoever is writing the code in Python has performance in mind. The problem I encounter

I agree with you overall but there's a huge caveat: you need to make sure whoever is writing the code in Python has performance in mind. The problem I encounter a lot is that many Data Scientists write unoptimized code that is taking longer to run than it should(ex. using for loops instead of numpy vectorization). You could have inefficient code in Rust as easily but even then the runtime difference would be the same.