I have not seen reports that pollsters are taking into account any sizeable population shifts in more than Southwestern states, so I wonder whether they may be missing it in other states such as mine, with Rustbelt population loss apparently reversing, but the myth of that loss still is so prevalent in many reports I see.    

I'd consider the DKos to be a "trial membership" sort of thing.  After a week or two, it should be clear whether or or not it's being well-run and on the up-and-up, or if there's some cherry picking going on with the released polling results there.

But you can correct for this. Suppose it is true that cell-only users who won't answer for an unrecognized number are young.  As long as you are reaching other young people (through landline and cell users who will answer) then you weight those responses more heavily in your final results because you know you are undersampling young people. You can look at the distribution of demographic characteristics in your sample and determine whther you are undersampling a part of the population.

This simple re-weighting of responses will work unless young users who won't answer their cell differ systematically in their political opinions from young users who will answer their cell or landline. Then the problem is more complex. But pollsters deal with correcting for systematic differences among non-respondents all the time. It's not a new problem.

That talk died down for a while with Obama's sweep of the Minnesota caucus and the crossover in the Wisconsin primary as well as with strong polls here -- until recently.  Now it may be that Minnesota and Wisconsin, the closest state last time, will be significant.  

But I bet that it will be more a "Minnewisichigandianio" -- ugh!  I.e., maybe Minnesota and Wisconsin but Michigan, Indiana, and Ohio as well -- i.e., the "upper Midwest" with a sizeable proportion of the population and, of course, the Electoral College.

Btw, I and others who study it seriously start with the census definition of the Midwest.  That also works fairly well in terms of somewhat of a legacy in shared histories, at least once they became part of American history in the 1780s east of the Mississippi and after 1800 west of it.  

So it includes the five states from the North-West Territory -- in order of Euro settlement (mainly French, as the Brits held them for only 20 years), that would be Michigan, Wisconsin, Illinois, Indiana, and Ohio (although Ohio, in typically Eastern-centric terms, calls itself the Gateway to the Midwest:-).  

And it includes seven of the states from the Louisiana Purchase -- also in order of Euro settlement, that would be Missouri (although an outlier in other ways as the only slave state of the 12 -- although all had slavery, illegally) and Minnesota, Iowa, Kansas, Nebraska, and South and North Dakota.

What is the Midwest?