MqPiffle states a rule I find myself agreeing with. The total basic amount of xp, a specific encounter yields, should be constant.
That means the model I proposed on page 3 isn't viable. Luckily DexCiscos model is sound and would yield the results I was aiming for - linking xp to risk taken. DexCiscos model would make me use low-level heroes to hunt darklings, and high-level heroes to hunt dragons, instead of just having a single stack of doom, where you have spectators earning xp.
Just splitting xp evenly would be a good start, but in my opinion, it would still favor the stack of doom strategy to heavily.
Besides the xp issue there are two other really big issues that needs fixing regarding heroes.
1) They break the economy. In my last game (Military victory on a small balanced map on normal difficulty in 24 years) I put the tax rate at zero in turn 1 and I never raised it, nor did I miss having gildars at any point in the game. It should not be possible to run a loot driven economy. At most, weapons and artifacts should be sellable for minimal amounts of gildars (or iron, or mana or crystals), i'm thinking 1 to 5 units pr. item at most, unless the item was specifically meant as a monetary reward (stuff like wolf pelts and spider eggs).
2) There is only one viable strategy for how to utilize your heroes, send them adventuring. Different strategies should be viable, hence there should be incentives for keeping your heroes in your cities. This issue is of course linked with the problem, that your cities aren't in any real danger from monsters. However, I personally also think that heroes should be able to contribute to your nation building when stationed in cities, to a much larger degree (a mage could add research when stationed in a city, a warrior could add xp to trained units, a defender could add xp to garrison units etc).