I think the key distinction is whether the conference presentation and writeup are viewed as a presentation of new results or as an exposition of work published elsewhere.
When you are publishing a research paper, the criterion for coauthorship is making an important intellectual contribution to the research. Everyone in this category must be listed, regardless of their role in the actual writing, unless they specifically decline coauthorship.
On the other hand, there is no need to include these coauthors on further expositions: if you go on to write a survey or review article on this topic, then the authorship should reflect who actually worked on the article, rather than who did the original research. (The same way if you wrote an exposition of some of Einstein's work, that wouldn't make him a posthumous coauthor.) Of course, the exposition needs to give clear credit for the original research.
The border between these cases can get a little blurry. If something could be viewed as a research announcement (where experts would be learning of the work for the first time), then I'd say that puts it in the research category, rather than exposition, even if the primary publication appears elsewhere. The basic issue is whether you could cause confusion. If your writeup may be cited as a research contribution, then you should list your coauthors.
If you are in a field like CS, where conferences are a primary venue for announcing research results, then you definitely need to list your collaborators as coauthors on the presentation and writeup. In math, things are a little different: some conferences publish refereed research papers in their proceedings, but there are also other possibilities. Some places, like Oberwolfach, ask speakers to write up an account of their talk, but this is not considered a research publication at all. In a case like that, there's probably no need to have formal coauthors, as long as the text itself is perfectly clear about referring to the research publication and author list. (When collaborators of mine have written things like this, I've been happy not to be listed as a coauthor on the resulting document, since that way I don't have to worry about the writing.)
However, the most important rule is not to offend your collaborators. Leaving out someone who feels they should be included is worse than including someone you don't think is necessary, so it's always best to ask.