I think it's a demonstrated fact that nobody cares about the out-of-the-box distro performance, because recompiling Ubuntu for your actual CPU (or anything newer than Opteron) gives large speedups on every program. Doing whole-distro PGO and LTO would be challenging because there's no one true workload upon which all users agree. An optimization for me might be bad for you. Also PGO and LTO will tend to think initialization and startup code is "cold" whereas in a thing like a shell pipeline that exec's many programs may be bottlenecked by those functions. If I profile a long-running daemon and rebuild libc with those profiles, it could make startup times slower for an exec-heavy workload.
