No, I think the smaller size is probably a non-starter for the intended audience. This is meant to be a tool that a lot of these folks are using for hours and hours a day so overall typing and reading experience is paramount, and meant to be a "selling" point.
Cost-wise, yes--we're still grappling with costs of the e-ink screen. It's the by far the largest expense in the unit. We've got a good lead that I unfortunately can't talk about at the moment (under NDA).
A cut-down Linux really is overkill for this, and isn't capable of making some of the other specs happen (battery life, availability), at least as far as we've been able to determine. Custom code is unfortunately where we've been forced to go.
Embedded Linux (eg any distro using busybox for glibc) is (imo) the appropriate stack to use for this given how cheap a microcontroller with an MMU is these days, and once you get rid of all the bloat, is capable of very quick boot times (availability), and if the device is off (and not just sleeping), then its battery life is similarly extended. (Don't let the poor battery life of Android phones let you believe Linux isn't capable of long battery life. Smartphones don't actually get to sleep while the screen is off.)
Remember, the device can sometimes have enough time to sleep BETWEEN keystrokes while typing. (The microcontroller can sleep then wake and power up peripherals with only a tiny, imperceptible delay.) This isn't your typical Linux setup. FreeRTOS is plenty. For the limited things it does, it wants to be a no-compromise device. There is no on button. There is no off button. You grab the device, it has an image on the display because it's e-ink and you start typing. Assuming you've previously connected to a remote device the system will have done what's necessary to keep that connection alive (invisibly waking as needed). Your keystrokes can go through right away to the remote connection. Your results will hit the screen right away too. The device is always on and it is always off, if that makes sense. There's more to it--bits of stuff and bother and implementation details--but I hope this gives you the idea.
Around 12" (in a 4:3 body) is about as low as you can go without messing with the typing experience much. The problem at that size is the available e-ink panels are S-L-O-W; no fast refresh models are out there.
(You might ask why we would even do the 10.3" prototype if we know that's too small for keyboard reasons? It's mostly for marketing reasons when we actually try to try to get attention for this project. The 10.3 is going to use a Thinkpad 701c butterfly keyboard in a custom body just for the "wow!" factor in showing the sorts of things that are possible. The 10.3 would actually be a killer end-product, but no similar keyboard is in current production so we couldn't actually make it).
Looking at all of the options a simple full RLCD would probably be best for us--not eink--but we can't find appropriate parts for the size and resolution.
RLCD is an interesting option, I guess the sacrifice there is readability angles and paper-like-ness?
What are you going to do about the actual cost of the device? You've already got an e-paper display, but you're also adding a 'laptop' style enclosure, real keyboard, and if you're storing SSH keys, you're really close to running a cut-down Linux system. That's basically a laptop, and you could get the same thing with, you know, a laptop. Have you thought about scaling down the display, keyboard, and enclosure to something palm-top sized?