There isn't a huge benefit to writing things like strlen in Rust. But a lot of libc is big, complicated things like the DNS resolution logic in gethostbyname. These are also where we find a huge number of vulnerabilities in popular libcs. The internal logic of functions like gethostbyname should be able to profitably take advantage of Rust's safety features.
That and people underestimate the value of the entire codebase being in a single programming language and environment. If your language continues to rely on C dynamically linked libraries there remains a barrier that your workflow can't be translated across. This is more pronounced in languages like Ruby/Python/Java ofcourse but it's still true for Rust to some extent. Having everything written in Rust (or whatever language you prefer) has compounding benefits when it comes to productively programming on really large systems.
LTO and static compilation are available to binaries built and statically linked with musl (it's one of the main advantages musl has over glibc, aside from being a more readable and cleaner codebase). And musl has done a lot of work to have correct cross-platform implementations. Granted whenever Rust adds a new target, it should be easier for things to "just work" when cross-compiled, but Rust is far less portable than C + a modicum of macros at the moment.
As far as I understand, the main selling point of Rust is safety, so if we don't get any safety, why bother? Am I missing something? (Other than having some fun with Rust, which is obviously welcome :) )