Slow startups are potentially a language issue (which can be mitigated/eliminated in various ways, especially for a text editor), but I hope you realize that comparing the performance of two completely different implementations doesn't say anything about the language choice.

If the neovim native LSP were a direct translation of the typescript implementation, then that would be useful to compare. However, I doubt that's what happened here since doing that would be a waste of effort. If anything, it's more likely neovim's LSP was designed with knowledge of some of CoC's performance issues in mind (since it pre-dates it).

As for youcompleteme, I don't have too much experience with it. I used it years ago and don't remember any performance problems, although that's anecdotal. Regardless, a completion plugin like that typically relies on a cache built from e.g. clang's compilation database feature, meaning that (after the cache is built) the plugin just needs to access that data structure in response to queries. The design of the cache and implementation of the interface between the client and the server are what will likely determine the bottlenecks, not the raw ops/sec of the Python interpreter.