It isn't so much that only actual problems are allowed, it's that actual problems have certain characteristics, and questions with those characteristics make better, more answerable questions. Here's a major difference:
When you have an actual problem, you are doing something, and things are happening as a result. There are specific symptoms that frame the problem. Hopefully, you've already done some trial and error and a little research, so you are working with something concrete. People can ask you questions to clarify what it is you're experiencing. They can suggest things to try to narrow down the possibilities. The problem is something that other people are likely to have run into and solved. So it has a good chance of getting an answer that is an actual solution, that will also help others with a similar problem.
When people come with hypothetical problems, it's often more of a thought experiment than something that people can answer from experience. There are no facts to work with, so the question tends to be overly broad; the answerer needs to contemplate all the hypothetical variations. The OP doesn't have the benefit of hands-on information to fully understand their own question. It may be hard to prove that an answer is correct, or that there is a single right answer. These tend to be more open-ended, theoretical discussions, which don't fit the Q&A model well.
So the intent is really to focus on questions that have the characteristics of actual problems more than a strict requirement that it be an actual problem.