"Can I Build a rocket to the moon that's powered only by saltwater in Transit Tycoon?"
Some questions are just flat out stupid. They should be closed, not because they ask whether an action is possible, but because they are stupid. They are so conjectural as to be functionally rhetorical. They are ambiguous, vague, and cannot be answered in their current form. Incidentally, that's a paraphrase of the text of one of our close reasons, Not a Real Question.
The thing is, this isn't a problem of questions asking whether an action is possible - after all, Can I permanently kill important people? is a perfectly good question - I'd go so far as to say it's a downright important question given the nature of the Elder Scrolls games and others in the genre.
Others, like Can I cure the Genophage without losing a Salarian? are asked after coming across a specific situation, and wondering if it's possible to change the outcome. Still others, like Is there any way to break a Protoss Sentry's Force Field? are about overcoming a challenge that seems insurmountable at first glance. In still other cases, asking how to reach a seemingly unreachable object can lead us to discover that the original asker is playing a pirated game!
This is a useful category of questions, that happens to contain some incredibly dumb questions.
We don't need a blanket ruling that they are off topic, we just need to use a little common sense and vote to close the bad ones. If a question is clearly based on actual gameplay experience, and seeking a solution to a real gameplay problem that someone faces, there's no reason we shouldn't be able to answer it. Questions that are based on stupid conjecture should of course be closed. Adding the word 'how' does nothing to improve the question in most cases and can in some cases render them meaningless, as in the case of questions about whether it's possible to accidentally sequence break or screw up quests by acting rashly.
One postscript here: We should be looking out for cases of 'Can I' questions where the asker is asking the wrong question. For example, with Can I secretly kill every guard in a town? the asker was asking how to implement what he thought was the solution to a problem - but it was a poor and unworkable solution when a much better one existed. These aren't bad questions, but they can be made into better questions by trying to get the OP to back up one or two steps to ask for help solving their problem, rather than help with implementing what may or may not be the appropriate solution.