I have actually already put some thought into it.
The thing that puts me off more than anything is actually designing the creatures and drawing their sprites.
A ‘cartridge’ isn’t limited to what can fit in progmem, it has a data area and can eat up the entire FX cart (minus the mandatory header data) if you really want it to.
Sharing data is theoretically possible, but more awkward because ‘cartridges’ can move around on the FX so actually locating the other ‘cartridges’ would become a problem. Hence it’s better to try to fit it all in a single game, which should be possible as long as there aren’t too many features.
I’m thinking all three would fit just fine as long as the battle logic and the scripting engine are sufficiently slim.
I have no clue what @bateske meant, but take a look at this:
If you have a middleman computer, Arduboy to Arduboy communication is possible over serial.
My one concern with this is “It doesn’t work well if the room is too bright”.
I’d be fine because I live where the skies are perpetually grey, but some people aren’t so lucky.
I’ve actually been thinking about making an RPG game for the FX (but I’ve not got past the world building phase). As a result of that I’ve been researching Pokemon Gen III sprite hacking.
If someone wants to make a Pokemon game and is willing to handle designing and drawing the monsters then I’m willing to help out with the engine when I’ve got less other things to worry about.
The overworld engine is the thing I’m most interested in because I’ve already researched it heavily.
The battle engine and level up mechanics are a bit more difficult because to make them decent you need some well-balanced stat equations (and I’m terrible at maths). For the most part you could probably just copy what Pokemon does though, or maybe use a simplified version.
At any rate, unless you only have a small pool of monsters it would be a big project.