If you plan to support new code base on old devices at all, development of a large project will result in hundreds of decision points where you can either have more features and faster or easier to maintain code on shipping hardware or better performance on discontinued devices. Just how much effort would YOU spend in the later, especially with a hard deadline coming up?
A new OS is also likely to create new demands on device drivers. How much support are you going to get from the manufacturers after they have discontinued the hardware, got out of an entire area of business or simply went belly up? Anyone who has a working knowledge of the chipset could already have left the company or be engaged on other pressing projects.
I think the most realistic solution is to release all available and legally unconstrained knowledge about the platform to community so that they can provide solutions like CyanogenMod as long as there is sufficient interest. In the meantime, try to treat free updates to discontinued hardware as a glass half full. The vendor has spent millions of dollars developing, testing and certifying it, with no commercial gains for itself besides reputation.