For example, advanced economies depend upon there being diverse and specialized professions. So if everyone worked in (say) construction, we'd all starve; but that obviously doesn't make working in the construction sector immoral. Even if construction work is widely regarded as permissible, there is no risk of everyone doing it, and hence no risk of disaster. Similarly for choosing not to have children. As these cases suggest, the relevant question turns out to be, not "what if everyone did that?", but rather, "what if everyone felt free to do that?" The answer to this latter question will often be, appropriately enough, "no problem!"
Other times, the "what if everyone did that?" heuristic serves to highlight a genuine moral problem, but one that can be equally well addressed from a straightforwardly Act Consequentialist perspective. This can take two forms:
(1) We often unjustly neglect the aggregate impact of small differences made to large numbers of people. So, iterating these effects can help to make them more visible, and bring us to see that actually each individual instance was more significant than we realized (i.e., more significant than was a competing, more visible effect on a single person).
(2) Sometimes decisions (e.g. buying factory-farmed meat) have what I call "chunky impacts", or threshold effects, whereby the vast majority of instances have no effect, but then a single threshold-breaking instance has a proportionately huge effect, such that the expected value comes out to the same (if you have a proportionate chance of being the threshold-breaker).
In either case, there's no special "collective action problem" that requires us to pretend that we're deciding for everyone. Simply assessing the expected utility of our individual action yields the right result.
Things can be different in threshold cases where you have background knowledge suggesting that it's (disproportionately) unlikely that you'll be a threshold-breaker. Voters in non-swing states may be in such a position, where the odds of their (presidential) vote making a difference are so slim that the expected value of their vote is negligible, even given the high stakes of a presidential election. (This will, of course, depend on the details!)
In this case, Act Consequentialism may recommend not voting. Critics then object that if every right-thinking citizen followed this recommendation, then even "safe states" would no longer be safe, with disastrous consequences! But note that AC's recommendation is highly contingent on our background evidence regarding others' dispositions. As it happens, we know that many people vote for expressive reasons, etc. If this were to change -- were we to suddenly find ourselves in a society of Act Consequentialists who don't get any intrinsic value out of the act of voting itself -- then the expected value of our voting would also change. For each right-thinking citizen who "drops out" of voting in a once-safe state, the expected value of the remaining citizens' votes increases, until you reach a point where the optimal number of right-thinking people are voting.
Of course, things can be trickier if the society falls into a bad equilibrium. If every right-thinking citizen drops out simultaneously, then subsequently there will be no act-consequentialist reason for them to start voting again, as no one alone has any chance of making a difference. (One can multiply examples along these lines, involving firing squads, etc.) The proper solution to this pure coordination problem is to build in some of the insights of Donald Regan's Cooperative Utilitarianism:
The basic idea [of Cooperative Utilitarianism] is that each agent should proceed in two steps: First he should identify the other agents who are willing and able to co-operate in the production of the best possible consequences. Then he should do his part in the best plan of behaviour for the group consisting of himself and the others so identified, in view of the behaviour of non-members of the group. (p.x)
In a world where all the right-thinking citizens are cooperative utilitarians, and some X% of them need to vote in order to avoid disaster, then a bit over X% of them should vote. (Perhaps each person would use a randomizing device to determine whether they fall into the voting group.) Meanwhile, in the real world, for as long as there are vanishingly few cooperative utilitarians around, those of us in non-swing states probably needn't bother.