Maybe you should pay to have built a batch of a thousand custom boards that are similar to, but not exactly the same as, the standard Raspberry Pi, so that they have a proper serial port etc.
You could also consider putting more than one ARM computer on each board (at which point maybe you want fewer than a thousand boards) and building an ethernet switch into that board, etc, and maybe you could even hardwire all the serial lines within the board, etc.
maybe. My partner has experience with that sort of thing, so yeah, that's an idea for a kickstarter or something. It would be awesome publicity for the sorts of people I like to sell to.
But from what I'm told, it's a fairly simple soldering job to put a serial port on a regular pi.
Personally, I'd want to add more ram if I'm going to bother with a whole goddamn board spin, and maybe remove the video hardware, but then, I don't know how involved that would be, and meh, I have a tonne of other work she could be doing now that she's working here that would probably provide a better return, and eh, the window for starting a custom spin of a pi board and having it done before the pi is obsolete is probably closing anyhow, so maybe next revision?
That's big problem with custom hardware... By the time you are done with all the Engineering effort and debugging, well, there's better hardware out there. Remember the neo? That's why you'd want to start with something mostly done (that was only recently released) and then make minor changes (and why removing video and adding more ram might be a bad idea?)
I would also need to figure out some automated bootloader setup, of course.