My $0.02 on these:
I am opposed to #1. Difficulties of "real" species identification and speciation over time aside, allowing different costs per specie is equivalent to allowing different laws of the universe per specie, which I am opposed to. IMHO, costs, phycis, etc. should be eqivalent for all species at the same phycisal locale, thereby providing a level playing field for evolution to operate. Note that I am *not* opposed to providing different physics, costs, etc for all species as a function of locale, via such mechanisims as the e-grid or via multiple connected sims ala internet mode. This is equivalent to providing different environmental niches, to which evolution can adapt. But IMHO, all organamisms within a locale should expereince the same costs regardless of species.
As an aside, I will also state that in my opinion, that fact that we must provide environmental costs at all, particularily as a mechanism for population control, is forced, artificial and mostly grounded in our limited computation resources. Costs provide a means for phenotypes to compete through effeciency, and thus for selection to operate along this dimension, but IMHO, effeciency relative to costs alone (as opposed to effeciency relative to environmental conditions such as tracking prey or navigating around shapes) is not the most or even a very interesting battleground and in a world with vastly more computation resources, we could very easily do without it. I would prefer to see competition for limited resources such as nrg, physical space, compition for prey, for specific habitat, effeciency in running, fighting, hiding, navigation, prey tracking and so on be the primary battlefronts and selection drivers.
One might argue that having no costs results in either unacceptable population explosions (unacceptable due to limited computation resources) or the necessity of limiting incoming nrg so severely that sim activity suffers and thus costs are something we can't do without today but I routinely run evo sims with no costs using shepard preditor bots as the sole means of population control. This works very well and has many advantages over cost-restricted sims I won't go into here.
A sim with an increasing CostX over time may be evidence of selection for efficiency relative to costs, but I think that is pretty much all it means. I would rather provide enviornments where complex organisms evolve with complex adaptations selected for complex beahviours necessary to compete against other complex organisms in a spiraling co-evolution that favors complexity over effeciency. A body maintence cost for example, is dead last on the list of things I want influencing selection of body size. I woudl much prefer preditor-prey interaction and so on be the primary selection drivers on this and every other morphological trait.
I have no objections to #2 and #3. I will add them to the list.