To be honest, I don't see any sensible argument, why creating device from scratch should be more fail, compared to Your ideas. At least, we have some predecessors (Open Pandora, Raspberry Pi, etc), while no one *ever* 100% reverse-engineered device like N900.
If, by some unbelievable miracle, You would have enough money to buy/reverse-engineer/whatever way open N900 "Stack" fully, it would be much better spent creating new device, running mainstream debian, archlinuxor whatsnot. For that money, probably 2 or 3 of them, in straight line
You're saying that developing own device is "we will bring You heaven" - to be honest, IMO ideas of fully opening N900/N9/whatever proprietary device You have in hand are even more unrealistic - basically, it just *won't* happen, ever. Probably, no matter how much time/money You spent on it. don't get me wrong - as You probably noticed, I love N900. Yet, projects like CSSU/kernel-power are best we can achieve. There is no feasible way to go through/around closed hardware-specific bits. Ask any developer with brain and common sense. Period. /Estel