This hypothetical game that I'm bearing in mind when I make these graphics. The basic idea is a Kingdom Of Loathing-like browser-based vaguely multiplayer game. Not exactly like KoL, nor an optimistically conceived "KoL... but better!" I am inspired by KoL, and I don't mean "I want to rip KoL off". I mean that KoL spurs me to be creative in my own way, that it serves as an example of what's possible.
The basic idea is that you, the player, is in charge of a guild of fantasy adventurers. You start off founding the guild with your leader character and gather more guild members over time. The central gameplay system is the "group of characters vs group of characters" battle as described previously. Outside of those battles there can be economic gameplay based around hoarding and consuming resources. Other gameplay systems could be "plugged in", but those two are what I see being where the bulk of the fun living.
The importance of the setting is how much it allows for. A fantasy mish-mash milleu allows for an almost unlimited variety of styles, characters and monsters. And it immediately lets the player know the sort of thing they should expect. Such as guilds of adventurers. The guild setting is far more important than the world though. It provides a rationale for managing a stable of characters, gives you an identity to become attached to and a good reason for your characters to seek out trouble.
Multiple axes of advancement are in my mind, what makes this setup particularly good for a game. You can level up and customise individual characters to improve their stats and unlock additional attacks. You can expand your roster of characters (who can then in turn be levelled up...). You can earn money to swell the guild coffers. You can expand and upgrade the guild house with new facilities. You can increase your stash of loot. You can complete missions to improve your guild's reputation and unlock more content.
And consider how these can interact - you take characters on a mission and they get stronger and earn the guild money. You spend the money expanding the guild house so that you can support more characters, or to provide facilities for those characters to use. And these interactions can be "squishy". You might decide that upgrading the guild is boring, so you focus on getting as many recruits as possible, just barely spending enough on the guild that they don't leave in disgust. Or you might really go in for pampering a few favourite characters with sweet bonuses from a nice guild house, and sweet loot made in your guild's armoury. And that's not touching on the economic gameplay, which can be used to enable guild progression or just an end in itself.
To enable replayability I think the KoL method is best; make play cyclical. I'm imagining a "valhalla" where legendary champions live. When you "complete" the game, your guild's leader (usually the character you started with, but there can be options and events that give you the option to switch who your focal character is) becomes a legendary hero and becomes saved in his current status to your account. And then you start again from scratch.
But as you play, you slowly build up a roster of legendary heroes in their "maxed-out" state in valhalla, and that's almost like a meta-guild for you. I imagine that this would be where player-vs-player competition would occur, though there's no reason there couldn't be a bit of single player content there too as a final tier. You could also get various benefits in the normal game from restarting, like access to new content, guild upgrades or loot. Perhaps the old champions could become a wandering heroes that you can meet. The possibilities are too wide to go over in any specifics here.
The currency of play would be in character actions. Like KoL (again) this would be a limited play-per-day game. Each day, each of your characters can do one thing. This will often be going on a mission (in which case you'll be using up several characters' actions to do it). For characters who don't join a party for a mission, they can be put to work at the guild. Maybe they can craft items. Perhaps they could be hired out as mercenaries, or sent to compete in an arena - simple things with results based on character stats/skills that don't involve active gameplay like battles do, although they could be expanded to minigame-status if needed.
Downtime skills could be a class of character skill that gvies them a special option when idle. A burglar might bring you extra money. A researcher might come up with a new spell that one of your wizards can learn. A chef could make delicious cake. This means that even when you have enough characters to form multiple parties, you could just select a bunch of downtime options for your characters and call it a day.
Lots of things I've not talked about. Interaction between players. What would matter from the economy side of things. Potential for expandability. Structure of missions and how they're dispensed. Inter-character relations absed on a few simple personality checkboxes. Special events and challenges. How collectability would be implemented.
But this isn't exactly a proper design document. Best to make sure of my big picture before I start getting all fiddly and specific.