The Trustwave "analysis" obviously has a severe bias and its sample size makes it a joke. It's unfortunate, because there are many legitimate examples that could have been used instead. Just a sampling from my changelogs shows tons of vulnerabilities whose fixes haven't been backported to earlier stable kernels.
The PTRACE_SETREGS race vulnerability in various incarnations goes back to at least 2.4, so at least 12 years of vulnerability (both on x86 and x86_64 BTW). FWIW, given the characteristics of the vulnerability, the constraints on it, and the extensive cleanup required to not bring the system down with it, it's unlikely to be exploitable on a grsecurity system with KERNEXEC/UDEREF enabled. If it were possible, a large infoleak of kernel .text would be needed (which we've hopefully eradicated via USERCOPY) and an additional infoleak or reliable address with which to store a ROP payload.