To answer the question:
Yes, it can be reliable, if it's done right, meaning that the wording of the questions has to be valid, unambiguous etc. No leading questions like "do you still beat your wife?", no suggestions of correlation as causation, like "as the sale of deodorant rises, so do the number drownings, should we stop the sale of deodorants?"
And yes, there has been done extensive research on fringe beliefs, for example about birthers and deathers (the conspiracy theory that Obama Care will kill the elderly). See google and wikipedia en.wikipedia.org/wiki/Birthers for starters.
"Polls conducted in 2010 suggested that at least one quarter of adult Americans said that they doubted Obama's U.S. birth, while a May 2011 Gallup poll found that 13% of American adults (23% of Republicans) continued to express such doubts"
So, you're right that it's disturbing
PS. I wonder, why would you assume that fringe beliefs makes the polling unreliable? Also, why do you think the polling IS unreliable to start with?
If sizable minorities believe in something, even though they are demonstrably untrue, is it then accurate to call them "fringe"?