I'm still going to to blame the sampling methods, as opposed to calling out "fraud."
At root, the methodologies aren't that different, really: to collect most (not all) of the data, the polling organizations have all still got to call folks on the phone.
So the question is: is there, in Nevada, some systemic factor that affects who you can call, and who is likely to answer, that was not accounted-for in the polls? For example, is there a "no-call list" law that favors land-lines over cell phones, or vice versa? Is there a particular demographic of voters that somehow are less likely to be reached by telephone?
I'm willing to bet that the answer to the poll discrepancy lies in that sort of phenomenon, rather than fraud.
An interesting and highly relevant example is the famous Literary Digest poll of 1936, that predicted Landon would defeat Roosevelt in a landslide.
The sampling methodology was very carefully constructed, but the problem was that the sample population was selected mainly from automobile registration and phone books. The surveys were mailed out, and results were drawn from those who returned their surveys, and the response rate was low. The combination of these three, led to a disastrously wrong prediction.
Pollsters have been trying to deal with similar problems in recent years. Lots of people refuse to participate, and many hang up when they recognize that tell-tale pause between answering and the time when the pollster hits the line.