1) The linaro code will deliver nowhere near what users are expecting from it. I believe that in their benchmarks, they disabled a bunch of hardware acceleration functions in order to rely more on the CPU to do heavy lifting. In addition, our GPU drivers (and many others) are binary blobs. You won't see significant changes on any device that relies heavily on vendor blobs.
2) The linaro prep work has broken a ton of ****. Tethering was busted for a day or so, and sometimes in the past few days something got broken that prevents a number of apps from pulling information from the internet (for example, Foursquare can't pull the list of friends' checkins.)
I'm fully expecting to get a GS3 next week, but I'll still have my infuse, which means I could do work on the infuse without having to take it back to a working state at the end of the evening, and might be able to help out more (or at least, worry less about breaking it in the process).. or at least more willing/able to be a quick test mule.
Now, #1 being said.. the linaro improvements still seem like an overall good thing for android in the future, since they're getting some changes integrated into gcc and the kernel to allow better arm optimization to be done by the compiler.. some people who have had good luck on nexus devices say their biggest difference has been battery life rather than raw performance. Which makes a lot of sense if you're using the hardware to its full potential rather than doing the same thing with more basic instructions..
But yeah, given the large amount of binary blobs in performance-critical and battery-draining code, it won't help that stuff until the toolchain is mature enough that samsung et al want to use it when they build their binaries.