First, to get the latter out of the way: Yglesias's publication has spent the past year or so breathlessly hyping Twitter controversies about the Democratic primaries, so it's nice to see Vox admit that this major staple of its political reporting is actually entirely irrelevant. But this point that Twitter is a dumb basis for political reporting isn't just something writers like me and Matt Bruenig have been saying for months and months - it's data that we've had for years, and it's mostly common sense for anyone who lives outside the elite media bubble.
From here, Yglesias makes his own contribution, arguing that Twitter isn't distorting a Democratic schism so much as inventing it entirely. "In contrast to casual sample taking on Twitter," he writes, "statistically valid surveys done by places like the Pew Center consistently show that Sanders supporters and Clinton supporters have similar views on the issues."
But there's a reason why Yglesias has to immediately add that "One can, of course, quibble with these findings." Consider this finding from the very survey he cites:
Clinton is closer to Trump on this issue than she is to Sanders, and only slightly further away from Kasich. And hasn't precisely this divide been at the center of so many Democratic primary controversies? Clinton argues that we just need to "save capitalism from itself," that her ties with Wll Street are no big deal, and that we mostly just need a more diverse cast of plutocratic overlords; Sanders declares himself a democratic socialist, argues for breaking up the banks, and refuses to work with super PACs. These aren't peripheral or superficial debates; they are substantive and at the center of controversy in the Democratic primary.
One point that Yglesias gets right, unlike pundits who are opportunistically hyperventilating about the so-called #NeverHillary campaign: most Democrats are obviously going to rally about the nominee. But this only reflects a "consensus" in the same way that polling on some of the issues reflects a "consensus": by virtue of framing.
If you set up a two-party system that forces Americans to make a binary choice between Ted Cruz and Hillary Clinton, obviously anyone who is oriented to the left is going to prefer Clinton. This is a comparative preference, not an absolute preference; all it tells us is that Cruz and Clinton are more different than Sanders and Clinton are. (The point holds for Republicns and Trump too, by the way - despite the #NeverTrump posturing, they will obviously rally around the nominee, just as Romney's totally antiestablishment, principled Tea Party critics predictably rallied around Romney.)
Similarly, if you only ask Americans "does the government have some vague responsibility to provide health care in some unspecified way (y/n)", obviously liberal centrists, progressives, democratic socialists, hardline Stalinists and third-world Maoists are going to tend to agree. This does not, when it comes down to policy specifics, imply any sort of actual, substantive, and meaningful consensus whatsoever; it mostly just means that a diverse political coalition rejects a major premise advanced by libertarian reactionaries.
Yglesias misses this diversity because he sees the right's "silent conservative majority vs. liberaldemcommiecrats" framing as essentially correct; he then reads the choice we are forced to make by the two-party system through the lens of confirmation bias, and concludes that the liberal Clintonites and socialist Sanderistas are basically the same. Twitter is a bad place to look to understand the Democratic primary, but the essentialisms of right-wing framing aren't much better.