A couple of points, to expand on the thoughts in this thread.

• As I'm reading it, that poll, at vote.bitcoin.com counts bitcoins, not users, right? So we can't tell the difference between 3000 people with one bitcoin each, and one person with 3000 bitcoins/addresses/something?
• The votes at that site line up almost exactly with the biases of the people who run the site. I'm not suggesting that it's being manipulated, but I think it's likely that it's better circulated in pro-big-block circles. The questions, also, are poorly written, and are likely to influence the outcome of the vote.
• The "other side" has their own poll using Coinbase's KYC to diminish the likelihood of multiple votes per person. Readers will note that many of my above criticisms also apply to this poll, and that it shows the opposite result.
• Hashpower is important, but it doesn't exist in a vacuum. Hashpower, like users, has an incentive to follow the majority of activity. If 60% of hashpower moves to a BitcoinY fork, but relatively few users follow, BitcoinY miners will be spending scarce resources, such as electricity, to mine a coin that has represents comparatively little real value. This, or other factors, could encourage existing miners to change their behavior, and even change the amount of hashpower existing in total. Miners are just as just as subject to incentives as other participants in the economy.
  Short term profit isn't the only thing that drives the behavior of miners. Ideology, long term gain, political power, externalities, etc. all play a role. I want to emphasize that this meme, "hashpower is the only really important thing," is probably oversimplifying a very complex . If heuristics like that were true enough to act on, we could throw out the entire canon of behavioral econ.