I guess alpha backer polls might be worthwhile on things actually in the alpha though.
Yes, I imagine after each alpha release FD email a link to an alpha survey for that release which the alpha backers fill in. Plus they clearly have a lof of back-end data they can analyse themselves re performance metrics, stability etc.
So that is what they are now 'polling'; alpha player surveys, defect lists and their own datasets.
The purpose of an alpha\beta is to prove\disprove the theories formulated during the design. So to actually change the design and hence those theories, once engaged in actual testing, would take a lot of work and would invalidate all the testing done to date and mean all the future test plans had to be revisited. In short, massive effort and increased risk for no tangible value. Clearly FD were happy with the design when they entered alpha, or they would not have entered alpha.
Now, should they get to the end of alpha and there is a lot of data (both empirical and customer survey data) saying that their design is in fact lacking and needs modification, they will have to revisit the design stage in some form. But even then, they would rely upon the first hand alpha tester survey data to forumulate views on possible ways to remediate the design, rather than open forum informal polls, which could be filled in by any Tom, or Harry.