Responding to the questions/suggestions so far:
Thanks for your help Tobe - it sounds like we should be able to sell a kit, maybe even with a ready-soldered-in module, without fear of jail time, except perhaps to Japan/Singapore and similar countries.
Built-in buzzer? Yes, we should do this.
Model memories? Possible. Only real problem is using up all the I/O pins with buttons for the selection. For those using a display, we can do a 'scroller wheel/roller' including 'press to select' using three input pins. For people who don't want a display, those three pins, and the two 'display' pins could be some sort of 'mode select' switches/jumpers - allowing up to thirty-two different set-up choices - but how would you configure the model memories without a display? There's also the possibility of using the main control sticks for selection/navigation with or without a display - but we'll need to think of a safe way to implement it that can't affect the transmitter when you're flying, and is reasonably intuitive for the users.
OLED display driver? Yes (optional). Providing it's just to display battery voltage, flight timer, and SIMPLE text-only displays, I think we can do it on the same chip. If you want fancy graphics then yes, a second Arduino would be necessary. There's also the problem of the OLED displays we use being perhaps a bit too small, and not visible in bright sunlight - but I've yet to find a cheap, readily available, alternative display. The two or four row, character-only, LCD displays, like I used on my single channel set, are readable in sunlight, but perhaps a bit too large, and old-fashioned for most?
Separate display/control module, that can be normally left disconnected? Yes good idea. We just need to program it so that the display/controls can be disconnected without upsetting the operation - then the same program can be used for both the 'built-in' display option, or the pluggable module version, or for those who don't want either. The connecting lead can't be too long, but people wouldn't want it much longer than a foot or two anyway.
Stackable expansion boards? Yes, if and when necessary. But I think we should concentrate on getting the 'base board' working first - I think most users will probably not want expansion.
How many channels? The ATMega chip has eight available analogue inputs, but if we want to drive an I2C display, that uses up two of the 'normally analog' pins. A battery monitor also uses an analog pin. So probably best to settle on 5 analog inputs for sticks/pots/sliders with a further 3 channels operated by switches - these could be two or three position switches, depending on how many digital pins we use for other functions. (If people are interested in the through-hole, easy to solder, chip version, that has two analog pins missing - perhaps we should forget about that option - at least to begin with?) Separate analog trim pots aren't feasible, but we already know how to wire trim pots to combine with the main 'stick' channels. There aren't enough direct digital inputs for digital trim buttons (but see below).
Buddy box? Do you want the transmitter to be the master or slave, or selectable? It uses up at least one extra I/O pin - two for master mode.
We need four pins to drive the module, one for the buzzer, perhaps one or two for buddy box? The ATmega chip has fourteen digital I/O, so with the extra channel switches, and other user controls, we're already close to using them up. Maybe we should consider some on-board I/O expansion to provide extra digital input and/or output pins? An alternative would be to use a different chip to the ATmega328 with more I/O built in - but I think that would frighten off many potential users who are comfortable with Arduino UNOs and NANOs. Thoughts?