2) you just simply don't model elections using regression analysis and a line of best fit. elections are not natural phenomena. a trendline in a graph is absolutely meaningless. smoothing things out is backwards. you want to amplify the noise, not quiet it down.
one should expect that election polling will fly around all over the place, not follow a defined curve. it may be true that a constant rate of change from now into infinity will give trump 100% of the vote. it's ridiculous to apply it as predictive at all. even over a few months.
but, that's my point. you can't do elections like you do physics. it's just wrong.
that said, i agree that he's right in pointing out that there's a lot of volatility right now. i'd just be more likely to argue that this volatility is careening towards a possible clinton landslide...
as to the point: is it too early to run out the clock? broadly, yes. on the emails, specifically (the point of the article)? no. there's lots of outreach she can continue to do, but i agree that she's better off keeping her mouth shut about that specific topic.
http://fivethirtyeight.com/features/election-update-its-too-soon-for-clinton-to-run-out-the-clock/