Ideally, Intel would wipe the floor with ARM on mobile and we wouldn't have to deal with this insanity anymore
The Copperhead kernel repositories (Samsung S4 and Nexus 5) for 3.4 aren't much different from what you've done. I don't think it makes much sense to invest a lot of time in 3.4 at this point because new Android devices are starting to ship with the 3.10 kernel and the lifetime of PaX's 3.2 LTS will be coming to an end. The main reason they're not yet public is because I didn't want my efforts to be judged based on an early proof of concept.
The OS will probably be alpha/beta-quality for the lifetime of 3.4, so significant improvements over vanilla while not matching a true grsecurity kernel with features beyond the PaX baseline and more security backports than Google/CyanogenMod provides is perfectly fine. Android is a lot more than the kernel, and most of the effort is going to be focused elsewhere in userspace. Spending less time on a kernel for old devices means spending more time on work that's going to be useful for a long time. Lots of the userspace improvements can also be upstreamed without too much of a struggle, which means spending more time developing new stuff instead of rebasing over and over and then migrating code to new Android versions.
I'd like to do a higher quality port for the 3.10 msm tree and closely follow the upstream 3.14 LTS, but I need to get my hands on a device that's using 3.10 first. This could be mature by the time 3.10 is the most common Android kernel and would hopefully last for some time.