One of the things I've been thinking about is a roguelite monster girl H-game with randomly generated dungeons. Fenoxo's Corruption of Champions already does this really well. There is no complex world map here - you head to a location and are presented with a random encounter, some of which open up further locations for exploration. This is a great example of abstract, extensible game design. If Fenoxo (or one of their writers) wants to add further scenarios, all they have to do is create the scenario and then slot it into the list of random encounters for a location. No massive rewrite of the whole game engine required (at least I assume so).
I really like the puzzle H-battles of Monster Girl Quest, but creating a whole Monster Girl Quest is a fuckload of work. I think an approach that could work is to create a more open game framework that can be extended easily with new encounters. Most roguelites/likes function on this principle - the dungeon is randomly generated each run from a series of building blocks.
Why not RNG the encounters like every other JRPG? Now, this is where, for me personally, Monmusu Quest: Paradox is weaker than Monster Girl Quest. JRPGs are a notoriously grindy game genre. The first time you encounter a new enemy is cool; the thirteenth time when you're just trying to run through to the next location, not so. Combat typically devolves to trying to bash the mooks off the screen as fast as possible. This is where Monster Girl Quest could get away with stretching out the fights with all kinds of delicious bondage holds. You only had to fight each monster girl once. If you really liked the fight (and Bad End) you could revisit it in the Monsterpedia.
One of the roguelite elements I think could work is a card-based encounter system. Camel uses a card-based mechanic to abstract dungeon-crawling in his Card Quest series (knowing what abstraction is and when to use it will save a lot of fledgling game designers/developers a ton of pain). I think this could be taken further to give the player some choice over the dungeons they romp in.
Hand of Fate is a good example from the mainstream gaming world. Here is TotalBiscuit's WTF is review of the game, which also gives an overview of the mechanics:
The player starts each quest with an equipment deck and an encounter deck. The dungeon floors are laid out as a random combination from both the player's and dealer's encounter decks. The player starts each run with just basic equipment. During the run they can buy or find better equipment - but only equipment cards that are present in their equipment deck.
I'm a big fan of strategy card games (like Magic: the Gathering) and the concept of tailoring your decks to "beat" the dealer's dungeon sounded very appealing. As it is, Hand of Fate doesn't quite play out like this. They have an interesting way of implementing side quests in that completing certain cards' scenarios will reward the player with a token, and this token unlocks the next card in the quest line at the end of the run. But the strategy element of the player customising their own deck is fairly weak - most of the time you run with the AI's default configuration anyway.
While I was playing around with potential stats for gamebook ideas, the three that seemed to have most potential were strength, agility and willpower. Depending on which the player specialised in, some monster girl encounters would be more or less difficult. As an example I thought up three lamia types - muscular, fast, hypnotic.
Muscular is slow, but if she ever gets her coils around you, you're never breaking out. Agile characters get the advantage as they just dodge her slow attacks.
Fast is... well... fast. But she's slender enough that her coils can be struggled out of. Strength characters get the advantage here.
And finally Willpower character gets the advantage vs Hypnotic Lamia. She's slow and weak, so she relies on hypnotising her opponent with sexy dancing so they don't resist her as she has her fun with them.
Coming back to a card system of dungeon layout. The player has some control over the encounters. If they have a low-Agility character they would want to customise their encounter deck so it didn't contain the Muscular Lamia that would likely be Bad End the moment they encountered her.
Or, and this is where it becomes really important for a monster girl sex game, the scenario of being dominated and squeezed in the coils of a powerful, sexy snake queen might be a major turn-on for that particular player, and they want her in the deck for the fun times when she shows up (and losing in these games is part of the fun anyway).
The reverse is also true. If the player is squicked out by vore, they'll want those nopey Eater monster girls banished as far away from their deck of encounters as possible. I think that's also a good thing for any H-game that's going to cover a very broad base of fetishes.
I think this concept could work really well. It's also really extensible. Dream up a new cool monster girl concept? Create a scenario card for her and into the game she goes without changing anything else.
Main problem is getting that framework right. And this is where I always screw up by making things too complex. I really want to make the fights more interesting than player hits monster for 5 damage, monster hits player back for 4 damage, repeat until someone reaches 0 hit points. I think Monster Girl Quest really got it right with the various bind and status attacks. Unfortunately, taking that too far can end up getting a little out of hand - like I did with this Arachne scenario I plotted out for a potential GameBook idea.
So I dunno.
Figuring out the player character stat line is important.
Figuring out what rewards/unlocks the player earns after each run is important.
Figuring out how the fights work is really really important.
I think the fighting is the key part. If I'm running mainly text-based to start, the other scenarios are multi-choice questions and stats tests wrapped up in sexy descriptive fluff. The fight encounters determine the mechanics. Is it sex-on-sex battlefuck, normal-on-sex like MGQ, or full-on puzzles with the player character trying to work out how to get to the next room without being caught and snu-snued to Bad End?
Figure that out and I suspect the rest starts to drop into place.