I can't find a place where it explicity states "one question per question" (in an official capacity, though we do have Framing (asking) good Questions for GIS Stack Exchange? linked from the help), but I have understood that to mean/be a response to questions that give all of their problems at once. In effect, the question is too broad.
One good example I have from recently is How do I create zones using location features? There are three distinct questions there. And all three have multiple considerations and possible approaches. A single answer to the asker's 'question' as a whole could (would) be very long. Another example that comes to mind is Network Analyst: Location Allocation for Model Builder where the basic question is 'troubleshoot my model'. We solve the first specific issue with the asker then moving to the next issue that came up. That turns into a running dialog rather than a Question and Answer.
In the example you cite, I think (and have understood from site help files) it's perfectly ok to answer 'I don't know how/you can't do it that way, but you could do it this way' with a full answer/explanation, even if it differs from the original question. This stems from the How do I write a good answer? page which states:
What, specifically, is the question asking for? Make sure your answer
provides that – or a viable alternative. The answer can be “don’t do
that”, but it should also include “try this instead”.
If there were no possible work-around, then by all means the answer should be 'you can't and here's why.' But in your example, you can't but you could do it this way would be a great answer to that question. Search results will be both questions and answers that have the terms - you could solve it without requiring another search. You did give the answer (just not the details), so I can see going either way - a new question or an answer there. I think the latter would be most helpful, or at least linking the current question to the new question (or one that already answers it, should such exist).
In short, I don't think it's a hard-and-fast or black-and-white rule. I think it's a case-by-case sort of thing, and some of them are fuzzier than others.