I've never been much impressed by the standard "counterexamples" to Consequentialism. They generally start by describing a harmful act, done for purpose of some greater immediate benefit, but that we would normally expect to have further bad effects in the long term (esp. the erosion of trust in vital social institutions). The case then stipulates that the immediate goal is indeed obtained, with none of the long-run consequences that we would expect. In other words, this typically disastrous act type happened, in this particular instance, to work out for the best. So, the argument goes, Consequentialism must endorse it, but doesn't that typically-disastrous act type just seem clearly wrong? (The organ harvesting case is perhaps the paradigm in this style.)
To that objection, the appropriate response seems to me to be something like this: (1) You've described a morally reckless agent, who was almost certainly not warranted in thinking that their particular performance of a typically-disastrous act would avoid being disastrous. Consequentialists can certainly criticize that. (2) If we imagine that somehow the voice of God reassured the agent that no-one would ever find out, so no long-run harm would be done, then that changes matters. There's a big difference between your typical case of "harvesting organs from the innocent" and the particular case of "harvesting organs from the innocent when you have 100% reliable testimony that this will save the most innocent lives on net, and have no unintended long-run consequences." The salience of the harm done to the first innocent still makes it a bitter pill to swallow. But when one carefully reflects on the whole situation, vividly imagining the lives of the five innocents who would otherwise die, and cautioning oneself against any unjustifiable status-quo bias, then I ultimately find I have no trouble at all endorsing this particular action, in this very unusual situation.
Eduardo Rivera-López has a fun new paper called 'The Moral Murderer. A (More) Effective Counterexample to Consequentialism', where he grants something like the above strategy for dealing with standard counterexamples to consequentialism. The lesson he draws from this is that an effective counterexample to consequentialism must (i) maintain the normal causal connections that we'd expect to hold in the described circumstances, and (ii) avoid undermining consequentialist-approved institutions.
Rivera-López's new counterexample appeals to (highly inconclusive) empirical evidence that the death penalty has a huge deterrence effect, such that each execution can be expected to save 18 lives. Rivera-López further suggests that it's plausible that a poor man in Oklahoma who killed a white woman in an especially heinous way could have a high enough chance of execution to render this murder a positive expected-utility act. So, should consequentialists all head down to Oklahoma and start murdering innocents, in hopes of being executed ourselves?
Well, no. For all we actually know, executions may not have any such deterrent effect. But that's ok for Rivera-López's purposes. We just need there to be a nearby possible world where this is (i) true and (ii) widely known (such that the agent can reasonably rely on it). So let's imagine that world. The pro-death penalty empirical research pans out, and everyone becomes well aware of the incredible deterrent power of executing murderers. Should the consequentialists in this world now become "high-impact murder-preventers", by themselves committing a single heinous murder and getting executed for it?
It's a funny question. I think the imagined world is still different enough from ours that it takes some time to wrap your head around the situation. (At least, I find it pretty hard to believe that execution could do so much good.) But if we really play along with the stipulations, then sure, I guess heinously killing one innocent and getting yourself executed could be a morally worthwhile act. (That's not to deny that it would be psychologically traumatizing, perhaps almost impossible, for any normal human.) That's assuming that the would-be murderers would not themselves have gotten executed, and that their non-murdering possible futures do not involve comparably bad actions, etc. etc.
Of course, it's hard to imagine that this could possibly be the best course of action available to a committed utilitarian. If he's sufficient intelligent and committed to be considering "moral murder", then surely there are even better opportunities out there for him to help people. Rivera-López responds that even if Tom the moral murderer is not doing as much good as he strictly could and ought to, he is at least doing more than most people, and so is (according to consequentialism) less blameworthy than your average person who is neither a moral murderer nor a professional philanthropist.
Here it's a little hard to assess how much good an ordinary person does through their everyday life, work and social relationships. (If they raise great kids, then that's a huge boon to the future. And folks at various tech companies have come up with innovations that improve the lives of many millions of people.) Being executed naturally puts an end to all Tom's efforts to improve the world. Is this one thing so much greater than everything else he would otherwise achieve? It's really far from obvious (unless Tom is in the last stages of a terminal illness. Perhaps it's just the octogenarian consequentialists to whom this case most strongly applies!)
Finally, it's worth distinguishing act and character evaluations in this case. Note that there would be something really disturbing about the moral character of someone who preferred to be a moral murderer than to realize an equal or greater good by more traditional philanthropic methods. Our ordinary failure to help non-salient distant needs, and to prefer to focus on more local concerns, is highly unfortunate but not positively malicious. But for someone to positively want to kill an innocent person, when the underlying philanthropic end could be better achieved a different (less harmful) way, seems quite dastardly. So even though the moral murderer does more good (by stipulation) than your average Joe, and hence is acting in a more worthwhile way, as an agent he may actually be more blameworthy, for his choice betrays an element of malice or otherwise bad will.
I think this analysis takes much (and perhaps all) of the sting out of the alleged counterexample. Yes, it's possible that a moral-murderer could be acting rightly, or at least no more wrongly than ordinary inaction. But we should still be very wary of the idea that such actions are normally a good idea, and so we may sensibly continue to endorse our intuitive moral disapproval of such acts in general. Moreover, even in a case where the act is worth performing, it's possible that the agent may reveal a moral flaw in their character if the salient harms do not bother them at all, or if they positively prefer to cause harm than to save lives by less harmful means.