The main issue is that these chips are made for cellphones. That means they expect to sell many millions of them per customer, and as new models come out pin-compatibility is pretty unimportant.
The ability for a small project like this to score really good ARM SoCs is slim, especially for the price range it's going for. I doubt you will see any system like this sold with SoCs that are competitive with the latest cellphones & tablets of the day, unless something changes in the market. No doubt if the RPi is successful (and it should be) I'd expect future models with better specs than current, but not necessarily the latest chips.
I have to agree with you on both of those points. However, as for the first point , sometimes OEMs do want pin-compatibility to try and cut R&D prices. One thing is designing a whole PCB from scratch, the other is redesigning an existing one, keeping the SoC area intact and only having to recreate traces for the new features to be implemented, right?
As for SoC prices, spot on. Low retail prices means using at least last-gen stuff. It has been like that forever, just look at what today's entry-level Smartphones are using (600-800MHz pre-Cortex cores). However, two good things: today's last-gen is about to be Tegra 2, Hummingbird, and generally speaking ~1GHz dual-core parts, which is nothing to sneeze at even a couple of years from now, if they make it into a Pi. And second, from what I've read on the IT news for these two couple of months, A15 seems to be much more friendly in terms of pricing, since you can basically use a single core design to just about everything from low-end to high-end devices, which is always good when it comes to reducing SoC pricing.
Now, one fleeting though crossed my mind. It may be from the cold I have, but how cool would it be to have a ~€50 dual-core, Pi-like, board capable of running Windows 8 for ARM? I'd be all over that one... hihihi Probably not available for at least another year, though, but dreaming is good, right?