I agree on the qualitative side - they're not doing anything at all it seems to make their sample representative in any way, and in the interview he said it was hugely skewed.
I understand the calculation in your post too, but most pollsters wouldn't go near those kind of tiny percentages. Even if the sample was perfectly representative, they'd have been laughed out of town if they'd said 0.6% of our survey have left the country in the last five months. But, in reality, most people running surveys know there's a certain amount of noise in even pretty good data, and I wouldn't rely on the maths with those kind of sample sizes (and yes I do agree that it's qualitative, but it's something that's amplified in smaller sized samples, beyond the simple margin or error calculation).
Still, as you say it's pointless. It's nonsense multiplied by rubbish, and yet still generated 100s of articles, which is rather depressing.