Haven't you more-or-less described OpenMoko's phone? It was hackable, expensive, aimed at developers, had poor take-up at first, and was critically dependent on a community of developers blossoming around it to turn it into something wonderful, yet out of either the maker's or the customers' control.
My impression is that this wasn't a success. Perhaps Nokia or Google could do a better job, but no sensible service provider would trust any software company to deliver on a multi-year roadmap of features, and no sensible customer would buy based on a promissory note of wonders to come.
Additionally, releasing a phone now that's designed to be good enough for use in three years time, in order to support the notion of continuously-added features, presumes that no new hardware will be needed (compass? HSPA+?), it requires the vendor to make the phone tough enough to survive three years of real-life abuse, and it presumes customers will be happy to have an unfashionable handset. A high-volume consumer market tends to work against all these presumptions, even if an extra (and up-front) investment in better hardware could be justified. Even in fiscally-prudent times, it's often still more cost-effective to buy a crappy handset every year and upgrade it, than it is to invest in a high-end handset and update its software for multiple years in the hope of keeping up.
While I applaud the ideal of high-spec, long-lasting, updateable phones, I just think the business case is no longer there for the general phone market.