This is such a great article. It's one of the examples that comes to mind when people ask me about great engineering blog posts. The TL;DR is the Go GC was causing some fairly large latency spikes under load. Since Rust doesn't do GC, because of its ownership model, these latency spikes didn't exist in their Rust implementation of the service.

Also, pflock you have a small typo there. You probably meant to say from Go to Rust, not Go to Go :)