Political polls are usually very short, many asking only the bare essentials about who you are, how likely you are to vote and who you'll vote for. You can do that in under a minute. And while polling isn't an exact science, you can apply scientific tests to them, including some of the things you've mentioned.
For example, we know that polls are about as accurate as they've been for over two decades - maybe a little more so. We know that telephone polls are generally more accurate than other methods, despite the fact that response rates have dropped off a cliff and overall access through that method has fallen. We know that polls have called the winner 4 out of 5 times, this close to the election period, stretching back to the 90s. We know that in close races, where the winning margin is 0-3 points, they get it right just 3 out of 5 times, while if the margin is 10+, they get it right 9 out of 10 times. There's a good summary
here.
So even it only reaches a sample of the population, that sample is not perfectly representative, and the measure being captured is not perfectly precise, we know how accurate it is. 1 out of 5 times it will get it wrong. When that 1 time time happens, it becomes proof that polls are unreliable, but when those 4 out of 5 times happen, it barely gets a mention. We're just very good at intuiting statistics!