The best way to think of it is that outstanding wounds on a multi wound unit aren't assigned to any particular model. When the unit takes more wounds you simply apply them, and remove models as necessary.
For hits that inflict multiple wounds into a multiple wound unit (Monstrous Infantry with 3 wounds per model, for example) the wounds inflicted by an individual hit cap at the max wounds of a single model.
Some Examples:
Bolt Thrower ( x D3 wounds, penetrating) into Ogres (3 wounds each)
First hit inflicts 3 wounds (kills an ogre) then penetrates and does 2 wounds against the next rank. Since that didn't kill a model, penetration ends and the unit has 2 outstanding wounds.
Then a cannonball bounces through, hitting 3 models ( x D6 wounds):
First hit wounds and multiplies to 5 wounds. that caps at 3 wounds, and kills a full model. In the next rank the hit inflicts 4 wounds, capping at 3 and killing another model. The final rank receives 2 wounds from the cannonball. One of the outstanding wounds adds up to 3, so another ogre is dead, and there's still 1 outstanding wound on the unit.
If there's any question as to when and how the outstanding wounds are applied (like why doesn't the first cannonball hit "top off" the outstanding wounds to 3 and kill a model, then the next hit (4 wounds, caps at 3) kills another, and the last hit does 2 wounds, so the model survives?), The rulebook says to "Add up all the wounds caused on the unit and then remove the appropriate number of models, noting any outstanding wounds." So like, Owain says, you add up all the wounds inflicted and remove whole models.
The first cannonball's wounds were enough to kill a full model, so it does the full 3 and the bounce continues. If the first hit only did 1 or 2 wounds, then the spare wounds would come into play and result in a kill, allowing the bounce to continue.