Not sure what's up with the author but from our brief interactions on X I have gotten an impression that they did not understand performance at a sufficiently low level and implications of using each language (and their compilers!) despite being adamant about the opposite, resulting in technical decisions driven by inaccurate conclusions. It was quite unfortunate and seemed like an emotionally-driven choice.

It's perfectly fine to use an LLVM-based language where it matters, but it's sad when C# could have been used to solve the problem without throwing the baby out with the bathwater.