There are a few edge cases that are generally considered ‘bad’ and best avoided in the study of voting methods. Things such as voting for a candidate actually reducing the chance the voter receives an outcome they would be happy with, or of there being a single person whose vote could alter the outcome of the election.
The particular passage they linked is talking about the spoiler effect, where a candidate close in ideology to another candidate causes both of them to lose and a candidate of a different ideology than the one with the most support wins. E.g. in 1912 Roosevelt ran against Taft, allowing Woodrow Wilson to win with 40 ish percent of the vote, despite Taft and Teddy have a majority between them and being much closer on policy.
Arrow’s impossibility theorem is a formal statement (and accompanying proof) that any ordinal system of voting, where voters indicate relative preferences between candidates, is subject to the spoiler effect. First past the post and ranked choice voting are among the worst in this regard, creating spoilers even in cases where voter preferences do not require such a condition to exist.
Other majority-rule systems such as condorcet methods are less sensitive to spoilers and, under certain models of voter preferences (e.g. smooth ideological preference 'i like candidates on the right and I like them less the more left they are), is immune to spoilers.
Ranked choice in the US is pretty much only instant-runoff (IRV).
All the votes are tallied for a voter’s first preference, if no majority winner then the ballots for the lowest-voted candidate are re-tallied for their next preference - repeat until some candidate gets a majority.
It is a terrible voting system, mathematically only barely better than first-past-the-post. The reason (in my opinion) that it has gotten some traction in the US is that it is the change that is least-threatening to the party duopoly, with the math telling us that is supports the same big-tent duopoly that FPTP does as an end-state.
A better form of it is the condorcet method, where candidates are compared pairwise based on preferences. E.g. on how many ballots is candidate A ranked above candidate B? The winner of such an election is the one who wins the greatest number of head-to-head elections.
STAR is interesting, but I personally find it too complex and it fails at preventing many different electoral pathologies.
Approval voting is the method I prefer (mark all candidates you approve for holding this office). Immune to spoilers, simple to explain, adaptable to single and multi-winner elections, and ensures that the candidate that has the broadest electoral support wins.