6 * Republicans, Libertarians, Democrats; 5 * Democrats, Libertarians, Republicans; 2 * Libertarians, Democrats, Republicans.
Liberians are eliminated (despite the majority of voters preferring them over the Democrats), and Democrats win.
Notice how when you use party names, it makes more sense? The Republican and Democrat voters are simply voting "anyone but the other major party". If they are neck-to-neck on primary votes, then the secondary votes of the minor parties start to be counted. This is a feature, not a bug.
Further, I don't think real-world voting would match up exactly with the above. There are people that would vote R/D/L and D/R/L, both of which are lacking from this example that would swing things away from the Libertarians.
I had a roommate who worked for a voting reform nonprofit and he said even gathering signatures for "simple" IRV is quite challenging. It's just too hard to explain quickly.
The plural of anecdote isn't evidence, I know, but: I had Condorcet voting briefly explained to me at a young age, and I grasped it immediately. I couldn't do the math, but it was conceptually really easy to grasp. I didn't fully understand the current US system until around age 20.
Unfortunately, HN readers are not representative of the general population. It's not that complicated, no, but I can see people having a lot more trouble with understanding Concordet voting than with understanding IRV.
* 8 people vote: A, C, B * 5 people vote: B, C, A * 4 people vote: C, B, A
IRV eliminates C, despite the majority of voters preferring C over B (12 to 5). The order of elimination means that significant secondary preferences are often ignored.