It would still be a self-selecting poll which they seem loathe to use for whatever reason.
See, what they've been saying all along isn't that they have concrete evidence of "fake" ballots, but that too many OS/2 people voted; that is, their population didn't behave randomly. Why? Because the proportions didn't agree with the supposed market share of the readers. Hence, the random sample.
But they still haven't said what they are expecting to find out from these things, so no one poll or survey is any more or less valid than any other.
For questions or comments, email the Forums Editor