For example questions like this that ask about if feature/algorithm X has an advantages/disadvantages over feature/algorithm y. Such questions encourage discussions and some of the questions and answers might be subjective. Should we consider them on topic or now?
Advantages and disadvantages are rarely very cut-and-dry, and as such tend to have some subjectivity to them. They seem to me to be highly-related to questions where the answer is "it depends," which we don't have a clear community view of.
So I would say that unless the advantages and disadvantages list is very broad, then these questions are generally subject to the "good subject/bad subjective" barometer.
In my opinion such questions are not subjective when the topic in question has a proved advantages or disadvantages. For example it is mathmatically proved that A* has an advantages over Dijkstra. On the other hand, such questions might become subjective were topic X doesn't have proved advantages/disadvantages and it becomes subjective to the observer interpretations. One example could be "What are advantages of Quick time events cut-scenes over playback only cut-scenes"1.
1 I just made this question up
I don't think you can ask an advantages / disadvantages question without it boiling down to a "Which tech to use" question (unless you are very crafty). These kind of questions are very situational and don't offer much useful insight to future visitors. They don't solve a problem.
Instead of focusing on general queries like what are the pros & cons of X, the question should simply be how could I get Y done, I used X so far and result are unsatisfactory because...
I think the rule of thumb is that if that once the question is answered, you should have your pragmatic goal accomplished and make real progress. Questions about pros and cons aren't going to help you complete any tangible tasks.
I remember asking questions like this once about Haxe and also about different ways of implementing the Events pattern. Reading the answers was an enjoyable experience but they were obviously not gong to help me make any actual progress. I still had to play around with Haxe and decide if I like it or not. I also never really had a problem with the Observer Pattern. I wanted to learn about alternatives implementations out of curiosity. The answer was informative and well written; I suspect it was not really useful to anyone in a bind.
I think SE really shines when it helps people who are in a bind and we shouldn't let people crowd with questions that are simply there for fascination and enjoyment although it sounds dreary, it helps keep SE lean and effective.