As for the PCBs, I would drop the proto-plus and offer a breadboardable DIP module and possibly the Retrino. I would make module ready to go with onboard crystal so it could be used standalone. In practice, there does not seem to be much demand for Mega format boards, so for a "Arduino" like board the Uno format might be better if that makes it cheaper.
The startup costs for PCB manufacturing and assembly are relatively low so leaving out a board won't decrease the funding goal or reduce cost of some other options.
Actually reducing the number of pins will not decrease the cost much either for the chip; this is the reason I did not keep the smaller chip. Reducing the MEGA to Uno size is just some PCB area and a few headers also not much in cost difference.
If possible, I would try to make a module that could be used directly in a retro computer. For that, it really needs 16KB RAM (or more), but necessarily onboard. I realise there is a big trade-off between on board RAM and providing EBI instead. Not sure what the best answer there is apart from "keep it simple".
In live one has to make choices and for the project I have chosen to go for a micro-controller with I/O as I think it would make more makers interested. Trying to combine both on one chip in this first project I do find too risky. But I will keep in my mind for future projects.
The current size of the chip is on a sweet spot of not much cost reduction anymore when making it smaller. Making it bigger though would increase cost almost linearly with the area so 16KB would cost considerably more. What may be a possibility is that if the campaign is really successful I could go for a smaller technology which can provide more RAM in the same area. But I'll look at this further when the campaign is launched.
Of course people could go crazy and make a setup where bank switching is implemented loading data into the on-chip RAM through the JTAG interface. Not very fast but doable.
I have some questions about how these boards are programmed, and the application tool chain. I helped create the stm32duino package for Arduino IDE, I think it would be useful to have an Arduino core package even if basic functionality.
For bare chip and also the breadboardable and prototyping board the chips could be booted from I2C flash where then the proper programmer can be used. Or it can be booted through the JTAG interface. I am using a Bus Pirate now with openocd program to do that but any supported openocd programmer should do and I don't think other JTAG programmers with other support programs would be difficult to support.
For the Retrino I do plan to start from the BlackIce (II) support programs. But the tools will likely be very basic and mainly focused on assembly programming. But as everything is open source I do hope the community will build on that and improve things.
Not much work has already been done on that yet but that is planned to happen in between the tape-out of the chip and the PCB ready (a few months).