I like the comparison but I think Scientist Time is an important missing part of the equation. (Also disclosure that I wrote my dissertation project in Julia, which made use both of multi threading and the ability to parallelize easily in an HPC environment.)

Not knowing rust, a lot of this material looks like it would have high initial investment (to learn rust) and an ongoing variable cost to implement (ie, figuring out all of these complexities in a particular use case). There is surely a trade off on accuracy but everything in science is about trade offs.

If you can publish three papers working in julia while the rust programmer gets through only one… we know who is going to get tenure.