Re: native vs JS. I'd say it depends. For certain implementations, such as operations on large arrays, native implementations can provide performance benefits (see https://github.com/stdlib-js/stdlib/tree/develop/lib/node_mo...). However, for other implementations, including common mathematical functions, implementing them in pure JavaScript can be more performant when running in Node.js, as you do not need to cross from JS to C/C++ (see https://github.com/stdlib-js/stdlib/tree/develop/lib/node_mo...). Lastly, even if functionality is backed by native code, would still need JavaScript fallbacks in order to, e.g., allow for bundling and use in non-Node.js environments.
Notable exceptions are R, Julia, MATLAB, Mathematica, etc. Namely, environments/languages which focus on numerical and scientific computing applications. One of the project's goals is to provide similar functionality on the web and in Node.js as those environments. Accordingly, in order to realize that goal, one needs to offer quality ("standard") implementations for doing mathematical operations (see https://github.com/stdlib-js/stdlib/tree/develop/lib/node_mo...).
https://blog.stdlib.io/reflection-on-the-metr-study-2025/