Useful strategies for answering "dumb" questions in a talk?

If somebody asks a dumb question, they're not going to feel good about the interaction whatever happens. If possible, get them back on the right track but avoid saying anything that could be interpreted as sarcasm and move on as quickly as possible.

The most likely thing is either that the questioner has missed something obvious or misunderstood something you said. So, in your first example, just point out that the rover is held there by the moon's gravity and move on. They're going to be embarrassed to have missed something so simple so it's important that you don't make them feel any worse about it; at least they'll probably understand the rest of your talk, now. In the far less common case of somebody making an assertion based on something that just isn't true, point out that the thing isn't true ("Well, the moon isn't made of cheese.") and move on. If they want to debate the point, offer to discuss it after the talk but don't let them derail you: everyone else in the room, you included, came for a talk about your thruster design, not an argument about whether the moon is a dairy product.

We are in full agreement that there are in fact dumb (or, more accurately, non-productive) questions, and I do not think that it is your responsibility as a speaker to make the asker of the question feel better about himself to the expense of the rest of the audience.

When this sort of thing happens to me, I try to answer accurately, politely and to the point as I see it, just as I would try to answer any other question:

Q: "I don't see how your the thrusters can possibly provide enough lift, given that the moon is made out of cheese!"

A: "I am afraid we have to disagree on the assumptions here. In my experience, the moon is likely not made out of cheese, hence this is somewhat of a non-issue in practice."

I would do the same if giving an interview talk. In that case, I would assume that the asker is likely just testing me. Rambling on, evading, or taking the suggestion seriously might actually be perceived as a negative in that case.

Since the OP specifically mentions the case of job talks and none of the other answers do, let me concentrate on that in my answer.

1) In a job talk, unless you specifically know otherwise, you should assume that everyone in the audience is someone who could have a direct hand in hiring you.

In the job talk I gave at my current university, a graduate student asked me a question about the arithmetic of Fano varieties. I began my response by carefully explaining what a Fano variety was and then quickly moving on to say that things like the circle method worked when the variety was "sufficiently Fano" in a certain precise sense. By the end of the day I learned that the person who asked me the question was not actually a graduate student but rather a youthful-looking tenured professor in algebraic geometry. In other words, she had forgotten more about Fano varieties than I would ever learn. Nevertheless, despite the fact that my answer was pitched a little too low for her, it did answer her question in a helpful and not condescending way, so she found my more-careful-than-necessary explanation more charming than offensive, and she joined me for lunch the next day. After I accepted the job, she quickly became one of my closest colleagues.

Corollary to 1: In a job talk, you cannot afford to answer anyone's question in a way which pokes fun at them with the hope of gaining points with the rest of the audience. The one person who got snubbed will remember that at the hiring meeting more than everyone else put together.

2) In any talk [i.e., a one-shot performance, unlike a course] you need to answer any single question in a way which keeps the overall talk on track. You don't want to spend more than a minute answering any single question, even if you know the answer and are happy to give it.

Thus you need to answer all questions in a globally efficient way. Since the question is about "stupid questions", I presume this means questions that you know the answer to. (If you don't know the answer to a question in a talk, probably the best strategy is to clearly acknowledge that in the moment you do not have an ideal answer, but that you'll think about it and be happy to get back to the questioner later on. It is tempting to stop short and wrestle with the question a bit -- this shows some positive traits, especially if you come out with the answer -- but it violates rule 2) above.) Moreover a "stupid question" is probably one for which the answer will not be enlightening to the rest of the audience, anticipate something that will come up later, or otherwise be worth spending much time on.

So I think the best way to answer a "stupid question" in a job talk is: directly, politely, and quickly. E.g.:

"Assuming I've heard and understood you correctly, the answer to your question is X. I'd be happy to elaborate, but I think it won't be so relevant to what I want to talk about today, so can we take this up after the talk?"

Note that this phrasing creates a polite amount of reasonable doubt that a stupid question was asked after all.

If all goes well, the questioner will drop the point and you can move on. Unfortunately, especially if the questioner is a high-status faculty member in the department, they may not want to drop it. In this case you should ask them to repeat the question, and you should take another crack at answering it, then say something like "And now I really feel like I need to move on, so that I can get through what I came here to say. But please feel free to talk to me about it afterwards."

Let me also say that I know a few "big dogs" who ask stupid questions that I have trouble believing are actually sincerely stupid questions. In other words, it is not unheard of that someone "plays dumb" during a job talk. I do not condone this behavior -- on the one hand, an interview is a two-way street and the would-be employers should be modelling their best future behavior just like the would-be employees, and on the other hand it is not so clear to me what constitutes a good response to such bad behavior so I'm not sure what they're hoping to gain (I hope it's not just trying to derail candidates that they have already decided they don't like: how ogrish) -- but I have seen it happen. But the above strategy is designed to combat this type of question as well: you want to give little to no offense but become minimally derailed.