There's a whopping bit of selection bias in that Gallup poll; it polls only people still living in Crimea. Some of those are Russians who immigrated in, and there's a big missing chunk of people who left because of the invasion.
It should be unsurprising that the folks who stayed tend to be the folks who support the Russian annexation. The opinion of those who fled matters, too.
That's definitely a factor, though I'd also add across all ethnicities. Remember we're looking back with hindsight. Imagine you're an ethnic Russian in Crimea in 2014. "Your" president has just been overthrown, there's been intermittent ethnic conflict rising, and now your territory just declared their independence. Your future is not looking peaceful, to say the least.
We can look at past populations though. [1] Ethnic Ukrainians have never been more than a sizable minority in Crimea. Their peak was 26.5% in 1970. And indeed there was a sharp drop in Ukrainians from 2001 to 2014 (going from 24% to 16%), though that was following a downward trend from the 1989 to 2001 census as well.
So if we take a high end estimate and say the real number was 25% Ukrainian, you'd still likely end up with a poll result of ~74% instead of 83%.
From the article: "The U.S. made a big deal about the rights of ethnic minorities there known as the Tatars, which account for around 10% of the population. Of the 4% total that said they did not endorse Russia's annexation..."
It's unclear whether they're saying 96% of Tatar's supported Russia, or "only" 60%. I'd assume the latter for balance, but the writing sure does suggest the former. In either case it's the same story. The reason for the absurdly high poll numbers overall is that ethnic Ukrainians boycotted the referendum.
It should be unsurprising that the folks who stayed tend to be the folks who support the Russian annexation. The opinion of those who fled matters, too.