So some talking heads--talking voices, I guess, although that sounds less clever--on the radio were defending the pollsters from the 2016 election

I didn't catch the whole segment, so I don't know if they distinguished the raw data (which were, I am reliably told, within sampling error of parity, which happens to align with the popular vote numbers we saw) from the reported results of 90% or 95% or 99% in favor of Hillary, whatever it was.

But that extreme error is what they were defending.

They went on about how people being polled were reluctant to appear too in favor of such a contemptible fool as Trump, lest it turn out that the pollster was a DNC employee testing the waters for an opportunity to dox a defenseless alleged bigot--even though doxxing wasn't so widely known a thing back then--or worse yet, the people polled felt that they would come off as bigoted even though they still wanted to vote that way.

So the pollsters apparently just struggled with trying to discern how the non-responses would have broken down by party lines, and how many dishonest results they got from paranoid voters, and so on.  

Those are fair concerns, but those are the concerns that define the pollsters' job.  This wasn't some historical anomaly that just coincidentally led them all right into the blind spot that corroborated every modern electoral conspiracy theory.  That, I would have excused if they had looked at the roughly 50/50 numbers and came back with a solid lead for Hillary but Don still making a competitive showing.

Ninety-plus percent, though?  

A landslide would be putting it mildly.  How much closer to a unanimous vote would it have to have really seemed to them before they stepped back and said "You know what, we better double check our methodology."  They wouldn't even have to admit to themselves any realistic bias like "Hillary is one of the more polarizing figures of the decade and we better make sure we don't undercount the people who hate her and still turn out to the polls."  All they should do is say "raw counts make it a toss-up, final analysis shows Hillary did almost twice that well and Trump got almost nothing; isn't that a stretch?"

No, I don't think it was an honest mistake.  I think it's a "mistake" they plan to make again, if our memory is short enough to let them, and they're trying to prime us with the suggestion that they're generally responsible people who happen to completely fubar their job but can be trusted to do better this fall even though times are even more uncertain all around.