So, I had an idea...
At the moment Vicious is seen as the pinnacle of difficulty, however there is a whole other layer of variable difficulty in the game through using unusual class-race combos.
For instance, monks rarely tackle dungeons with may debuff monsters. Elves are rarely berserkers, halflings are rarely spellcasters and few bother finishing most dungeons with more than a few classes. Certain gods are better with certain classes and races, as are certain items, and some dungeons are full of meat men and goos, or glass cannons.
Badges reward unusual feats, but don't cover as much as they could; once a player has barreled through a dungeon with it's most suited class, completed all quests and gotten all badges, why go back? The goals of the game are fixed, not dynamic.
Some players have been setting unusual and difficult challenges for themselves to alleviate the boredom between updates, to great success and amusement of the community.
The issue is that only a few players have the attention span these days to create and pursue their own goals like that, not to mention the fact that coming up with these challenges requires inspiration and knowledge of the game that not many DD players will have.
Given that badges and completion stats are already collected by the game, is there any reason why that data could not be used to figure out which race and class combos are the most and least used on certain dungeons?
If the game can track what race and class you pick the most and least often, it can provide rewards for trying others.
If it notices you have completed a dungeon with only a few classes, it can provide incentive to finish it with more. If across all games of DD people are using a certain class with a certain race, then a reward can be set for giving something else a go. If certain items aren't being used with certain classes, reduce the cost for prepping them with that class.
I think this would reveal a very large amount of undiscovered replay value and difficulty in the early dungeons. Difficulty is relative for each class and class in each dungeon, and leveraging that would help the progression and user-friendliness of the game.
I also think it would greatly reduce the feeling of 'exploit redundancy', namely the point when you realise that one class and race combo can stomp everything, and you lose the urge to try other classes. This was something that Lujo held close to his heart, and i think applying diminishing returns to exploited classes would give players a whole new lease on the game.