How? I made the assumption least favorable to Russia's position by assuming that that 100% of people who did not vote would have voted against the referendum. Russia got 97% yes with 83% turnout. Gallup got 82.8% agreeing that 'the referendum represents the will of the people.' Factor in 17% who didn't vote in the official referendum as no's, and you get pretty much the exact same numbers.
Yip. Sampling size can be counter-intuitive, but 500 for a population of 2.3 million is going to be well more than needed. For contrast the typical election poll in the US generally has from 500 to 2000 people, for a nation of 340 million.
Here's a calculator [1]. To get a 95% level of confidence, they would have needed a sample size of 217 people. This is assuming a population of 2.3 million and a response distribution of 83%. If the response distribution was 50% (which maximizes the necessary sample size) they'd have needed 385 people.
The referendum that was conducted by illegal occupation authorities obviously can not be trusted. So those numbers are made up propaganda.
With 500 people the sample size needs to be very carefully selected to eliminate bias and it is really easy to influence the results however way you want.
Gallup is more or less synonymous with careful polling. However, the final thing I'd add is that it wasn't just Gallup either. Russia claimed a total of 82.9% yes (if we assume 100% who did not vote would have voted no), Gallup found 82.8%, GFK (German polling firm) found 82%. The only 'outlier' (so much as 5% can be called an outlier) is Pew who found 88%.
If any of these organizations' operations had been impeded in any way they would have made a big stink of it in their reporting. US and German organizations were not there to support Russia's claims, but to challenge them! But ultimately everybody's numbers ended up the same.
This alone invalidates any of your claim that there is anything fair and accurate about the polling or election numbers.
It is as if you want to believe (and spread) fake things.