Regarding the shortcomings in polls, we have alternate sources of information about the current status of campaigns:
1. betting markets - these actually long preceded scientifically-valid opinion polls. The obvious question is: on what are the betting odds based if not scientifically-valid polls?
https://www.aeaweb.org/articles?id=10.1257/0895330041371277
2. the stock market - in close races, they move after the vote is counted. But, if the outcome is pretty much a sure thing before the vote is counted, they move before the vote is counted. Admittedly, this source of information is difficult to interpret in isolation, as stock markets move all the time for any number of reasons.
As to whether the polls are “pure BS,” they’re better than flipping coins, but nowhere near as good as they’re supposed to be given probability theory.
In 2016, for example, many of us who were hoping for a Trump victory, looked “into” the polls and saw a drift of Gary Johnson voters toward Trump and thought that draft would continue through the days immediately preceding the vote. We also saw that a lot of undecided minority voters expressed values associated with Trump; and, we thought in the privacy of the voting booth, they’d vote their values and not their skin color. (Hey, that’s one of the reasons we like voting on election day and we like the secret ballot.)
Polls have always been subject to confounding factors, but they used to be more reliable when we still lived in a society with a clear, shared moral code that placed a very high value on truthfulness, in which people mostly trusted the media and government, and in which voting was still considered a high civic obligation, at least among the broad middle ground of the culture.
All of those factors have been severely battered. People used to be excited when they were contacted by a well-known pollster; today, many routinely lie to pollsters. And on top of that, the mass adoption of cellphones and the use of call screening has made it increasingly difficult to reach people. Assembling a statistically representative sample has become remarkably difficult.
The point is, the reliability of polling isn't just a matter of gamemanship and push polling, although these remain factors. The bigger challenge is the highly competitive arms race among polling organizations to develop representative sampling models and statistical adjustments for the known confounding factors. Honest polling isn't irrelevant -- but you had better understand the methodology and the assumptions that are being built into the model. The genius of any election cycle will be the pollster who gets his "secret sauce" closest to the mark.