The "needs work" and similar judgments come from the Area51 stats.
We've pretty much always had a steady supply of questions which aren't easily answered, perhaps due to relative obscurity, or lack of OP responsiveness, other reasons.
These occasionally get a great answer, but most of the time they just join the pile and sit there. This drives the ratio
(answered questions)/(total questions) down.
Back during the beta there was more reason to care about this, as the site's success was presumably judged based on this stat, among others. You can actually see some of our efforts to stay above the "acceptable" threshold by closing obscure questions as too localized, organizing answer parties, etc:
Graph from stacker. I was unable to find a way to make the site you linked graph data from further back.
Note that the sharp drop near first intersection with the "acceptable" threshold is a glitch; a similar artifact can be observed in the data for other SE sites.
Once we graduated these stats became less important, and less effort was spent maintaining them.
Unlike programming questions, many blender questions can be pretty completely covered in a single answer (e.g. "how to change the theme?").
There are definitely many exceptions to this, but I think it isn't too surprising or concerning that the average blender question receives only one answer.
I don't think it's anything to worry about, because there are plenty of examples of slightly more open-ended questions being positively showered in answers (e.g. things like "how to model <simple shape>").
Most questions just don't readily lend themselves to such special treatment.