I tend to like Toady's way of handling it, with the current gobbos as just a placeholder. Between starvation, constant warring and some border constraints, it should hold pop size in check slightly for now. Or at least the places where the huge pop blobs can be found. It's kind of nice to have a place where you really have to try hard to genocide a people.
I'm hoping that the caravan arc will put the goblin's focus on snatchers and slaves. With the snatched/slaves being the only ones that can do "drudge work" (ie: providing all the food and labour that the civ needs to survive) and goblins being only 1/4-1/8 as good at basic labours.
It would still allow goblin centric societies to exist, with huge requirements of goblins to keep a few nobles and military elite going. But more "successful" civs, with heaps of snatched/slaves, are the only ones that can really get the goblin warmachine running (heaps more free goblins to be militarized, fed and armoured by their slaves). Non-slave gob civilizations may have more goblins (less war to kill them all off), but a slave-keeping civ would be terrifying in war (but with high attrition due to it's constant fighting of them).
It'd make it a definitely evil civilization, but with limits and an archetype for different styles of it. Between resources available and the ease of their use (needing four times as many people just to feed your military and themselves is huge) we could end up with a plethora of different goblin "styles", but still tied firmly into the gameworld and it's mechanics.
This could be done with skill caps, a basic output divider for basic labours, or a tag such as [DOESNOTWORK]. Maybe people enslaved could also get a tag [DOESNOTFIGHT], so that they're skill capped or poor at it, giving a huge difference in abilities between goblins and those that work for them.