This is an interesting report. Four polling organisations work with exactly the same raw data and come up very different results.
While this is nominally about Trump v Clinton, it does give some insight in how pollsters got things so wrong in this country. If there's a wide discrepancy on the data, before we start talking about sampling and whether questions are neutral, it's a wonder that polls aren't even further apart
I think it's very easy to draw a line from close in the polls to better ratings for cable news networks.
It really isn't in Fox, CNN etc best interest to predict a landslide victory, for anybody, certainly not in September.