@DuckRaGod I love it when people say something is impossible, only to later witness a group of weirdos do the very thing.

Count me in the willing to help column (even though I know next to nothing about GameDev). 😅

Yeah it's a nice idea but...have you worked with people? 😄

That's never going to work. Making a game with one lead developer is hard enough. A game by consensus? Forget about it.

Here's an idea:
A really simple bot that users can talk to via the fediverse and will spit out responses if talked to, like a roleplay bot. I suggest this because it probably wouldn't take much time to complete basic functionality, being purely text based, but could be easily expanded to make the gameplay more interesting over time. 'Multiplayer' could be done via a command like '@roleplaybot I attack @Stardust' (parsed by splitting/exploding the string) and having some kind of simple database storing HP and such.
Then if that works, maybe a more ambitious version of the game that actually displays images and such, perhaps 2d.
Example commands:
@roleplaybot generate me a dungeon/monster/goblin
@roleplaybot role dice d6
@roleplaybot DM a session between me and @Stardust
@roleplaybot Set my class Rogue Level 1
@rpbot Travel up stairs

Could be a lot of fun. You'd probably be looking at using something like javascript or typical webserver backend like php, ruby, some of the already existing fediverse code you could just start modifying. Take one of the existing scripts that do something on an activitypub response and include the roleplay bot's stuff. Nodejs seems like an obvious choice for a game + webserver but I'm not sure how much activitypub coding exists for it, haven't looked.

reminds me of https://github.com/a-little-org-called-mario/a-little-game-called-mario

I don't think it's that crazy - and it'd be fun to collaborate on something low-key, and maybe it grows into something?

This sort of thing can be done, but it takes a lot of work to coordinate everyone. Try to keep it as small as it needs to be so it doesn't spiral into an unfinishable procedurally generated VR MMORPG dwarf fortress / factorio pyanodon / no man's sky hybrid with AI NPCs, puzzle platformer elements, and ActivityPub integration.

You can look at Thrive, which is an open source C# Godot game aiming to be a scientifically accurate Spore. It has a really huge scope and has been in development for more than a decade. But at least they are working on it mostly one stage at a time, so it starts as a relatively small but complete game and grows from there.

And if you don't have a specific idea in mind, you might be better off contributing to existing open source Godot games. At the least look over existing projects so you're not duplicating effort. You might find one that you can fork into something new.