I take it you've never worked on a large code base with a large team? Even ignoring the actual task of rewriting the logic in a different language with differen

I take it you've never worked on a large code base with a large team? Even ignoring the actual task of rewriting the logic in a different language with different idioms, the change is going to be a massive hit to productivity for many months as people learn the new architecture of the rewrite, the tooling, and idiomatic Go. For a project the size of YouTube, that will be millions of dollars of engineering hours and weeks/months of lost productivity for an unknown gain and almost guaranteed bugs. It's a terrible value proposition so it's better to squeeze every last drop of performance out of the code base you have, which at the scale if this project includes paying engineer(s) to work on a completely new runtime.