Isn't this a fundamental problem with consequentialism? If you can convince yourself of a serious enough consequence, you can do whatever you want. Steal, hurt, kill. As long as your actions lead to a .000001% chance of preventing Skynet from taking over in the year 50000 AD, any means justify those ends.

How is that something that can be seriously engaged with? Other than trying to talk some sense into these people about limiting their EA calculations to a reasonably observable time-space continuum?