Our nodejs + python backends are all: multiple processes -> async/await nodejs/python event loops (used to be frp/streams) . It's good to solve event loop bottlenecks, but more from a individual query's latency perspective, not a machine utilization perspective: processes take care of that. Go seems great for hiring ambitious CPU coders, and Rust for ones making it safe, but the stdlib in both seem not there yet, so for most teams, the perceived increase in productivity is an illusion as it's wasted on building unnecessary things. (We actually went back to Python for web just so we can use Django, as the Node world is still too manual.)

Conversely, anything 'heavy' is probably too slow in Go/Rust too! We use GPUs for anything data-intensive, and the Python ecosystem is basically the best for that, similar to C/Java in earlier CPU eras. (I can see that changing in 2+ years as Arrow progresses and more investment happens.)