Other than consistency with coding style, I've never got the appeal of camel or pascal case. Is it a file size concern, maybe? I guess an underscore could take up additional space when used across a codebase, and maybe this has been held over?
No surprise to me. CamelCase can be ambiguous when it interacts with other capitalization conventions (acronyms, never mind words like a and i)
Why would snake case be faster than "kebob" case (this-is-an-example-of-kebob-case)?
Underscores really leave spaces as the main delimeter; i.e. when you see e.g. winsock_ctx, your brain is really keying on the space rather than the line. Compare that to winsock-ctx, where I notice the line more. We're good at reading spaces, both because we've been trained to, and because they'll always stand out more than using one squiggle among a bunch of other squiggles. Just my guess.