According to Michael Adams, an English language history professor who spoke to Smithsonian, we came to collectively accept "sherbert" because that's how we want to hear it pronounced. In short, there's an inherent habit among English speakers to unconsciously rhyme syllables that look like one another, even when it's incorrect.
"Sherbet is begging to be pronounced Herbert on this 'principle'," he said. "It isn't a type of systematic change in language, but a lexical change." Add to that the fact that people have been referring to the colorful scooped snack incorrectly for generations and inadvertently passing down the tendency to add an "r" for generations.
Another especially compelling theory blames the whole brouhaha squarely on a hit song from the Big Band era called "Shoot the Sherbet To Me Herbert," which hit radio waves in 1939. In the title, sherbet is spelled correctly, but in the song, it's bent so as to rhyme with the man's name, and voila: the sherbet/"sherbert" dilemma was born. Well, maybe.