Because Cairo was an attempt to abstract drawing operations over multiple heterogeneous drawing devices. In theory, Cairo should give you accelerated OpenGL rendering, PDF rendering, GDK rendering, PNG rendering, Win32 rendering, etc behind a "convenient" PostScript-like interface. Thus, when these devices gets new features Cairo must also be updated to exploit these new features.

That was the theory behind Cairo. In practice it never worked out and Cairo wasn't able to fully exploit hardware accelerated rendering. Cairo was very complex so it wasn't easy for the maintainers to keep all the various backends up to date.