I would like to distinguish three layers here: The level of personal input, the level of knowledge and the voting/commenting/deleting behaviour how it should be vs. how it is actually (sometimes) handled.
personal input vs. depth of knowledge
To highlight what I mean, I want to use the example of a first-year undergrad essay compared to a master's thesis.
On a first look, they write very similarly: They write on philosophical questions, using lots of references and remarks.
But they have different tasks: The undergrad essay is supposed to show the student is able to use literature, to paraphrase and reference correctly and get what the text is saying. Critical acclaim is only but a small fraction and the difference between distinction and (very) good. So they basically first and foremost have to show that they can write about philosophy in an academic style.
The master's thesis, on the other hand, definitely has to use literature, paraphrase and reference correctly, and get what the texts are saying, as well. But it also has to substantially go beyond (existing) literature, make good points against it and develop own positions. It should not only develop a position towards the existing literature, it should develop an independent and original position, highlighting the differences to literature. This is doing philosophy and critical for a master's thesis (read: at least should be!, latest here is PhD...).
Of course, this is a continuum and there are borderline cases, but considering the design of StackExchange, any site in the framework should methodologically stick to what is asked from a first-year undergrad, as it were.
Problem: Level of knowledge
Some of the questions are going quite into depth or broad enough so that they can hardly be answered by an undergrad (which is/may be good). This means if there are people able to deliver a satisfying answer (content wise), these people are either very deep into the topic and can reference off the top of their heads (perfect case), or know the answer, but would have to put a lot of time and effort they do not have in order to find references to support the knowledge they gained over years (this is bad). E.g. this recent answer took me some good hour to complete, although I worked on that topic quite a lot and had electronic versions of the corresponding sources at hand.
So basically, we should give both knowledgeable answers with a lot of background in order to answer at all, but at the same time invest the effort a first-year has to, explicitly referencing particular papers and/or books with pages, without adding a lot of personal opinion/flavour/content. This is a lot to be asked, which I can tell as there were answers I had to invest several hours into (even without consideration of much secondary literature).
Result: Lax/forgiving behaviour in direct conflict with the rules
So we basically are caught in a dilemma: Either we are rigorously downvoting/deleting/negatively commenting valuable, correct answers (possibly losing this content and the user), or we are generous and cautious, editing in sources (suggestion by @Isaacson) or simply silently accept that it is asked a lot in some cases to answer the question in a way that it should be.
This harbours the risk of a slippery slope: Where do we stop? And this is exactly what the community as a whole has to dynamically decide on in every single case by votes. I personally, as a moderator, would condemn some of the highest voted answers as bad ones. But it is not my personal decision, it is a social one.
My feeling is, though, that the more sensible or knowledgeable a post reads, the more forgiving the community is regarding sources. I do not think this is a good thing, despite being understandable.
I would like to see a conscience of the continuum between common sense/knowledge (e.g. logic) that probably must not always be sourced/referenced and more uncommon or debatable ideas or propositions that gradually more urgently demand sourcing. In general, if I have never read a position or know that it is a minority position, I should ask for sources and/or downvote - given a certain level of knowledge in that matter (which leads us to the next problem...).