Citations and external material serve two purposes, to support claims and statements made in the post, and to allow easier follow-up reading by the readership. At the end of the day, we should be willing and able to support the validity of any statement we make in an answer, so we should bear that in mind when we write answers. For example, one could say:
DNA is a nucleic acid composed of nucleotides.
Now a citation isn't necessary, because it's not really a contentious point. However, a link to say, the wikipedia article on DNA, would support that claim while also producing material that readers can use for further information should they want it. Therefore, a reference is often not necessary but would be beneficial.
One could instead say:
DNA is made of balloons.
This statement would benefit from having a citation to support that claim, because it is likely to be contentious. Obviously there is a scale, this is not a binary problem, some points will be more contentious than others. Ideally the person writing the answer would give a good reference to support this statement, such as a proper scientific paper, as a proactive referencing measure. However, often such contentious statements are poorly supported, or entirely unsupported statements. This will likely lead to comments along the lines of:
WTF?! DNA is not made of balloons! Please could you provide a reference to support your claim?
As a result, the provider of the answer should either edit to remove the claim or provide supporting evidence; reactive referencing. If someone writes something that they are not willing or able to defend or support then they should remove that text from their answer (or it should be removed for them).
As a general rule of thumb, I would say that referencing is not always necessary but we should
i) Always produce answers that could be supported if requested
ii) Use a generous serving of external links to help the user find definitions, fill gaps in knowledge that may help them understand the answer, or allow them to follow up with further reading
iii) Try to support any material that is likely to be either contentious in its validity or not well known with proactive referencing
iv) Provide reactive references to support claims if other users call you out on them. If you can not provide a reference you should edit or delete the material to make it fall in line with accepted or supported consensus
So while we should be striving to be proactive in our referencing, we should at the very least be willing to be reactive.
What to do about poorly referenced answers?
i) Comment and request references, ideally drawing attention to specific points of contention, or areas where the readership would likely benefit from external material
ii) Add references yourself to other peoples answers if you think you can provide the material needed. For example, you might see someone talking about the mechanisms of evolution, so you could add the links for the Understanding Evolution pages
iii) Downvote and tell the user you are downvoting due to the lack of references, again being specific about details you want supported
iv) Flag for moderator attention so they can add this banner which will warn other users that material is disputed.
Some of the information contained in this post requires additional references. Please edit to add citations to reliable sources that support the assertions made here. Unsourced material may be disputed or deleted.
Note that comments of "could you add some references to your answer" are useful, and indicate that the answer is generally in need of supporting material, but we also should try to be specific wherever possible, e.g. "could you add some references to your answer, especially to support your claim that DNA is made from balloons"
Unfortunately, there's not a lot else we can do about poorly referenced answers. If the writer of the answer does not want to provide references we can't force them. I've started a discussion thread on counter-measures.