I don't know much about the history of IRV in Australia, but the history in the United States has been nothing short of disastrous. It's only been used a handful of times, and yet there are multiple high-profile examples of major failures. Two of them are the 2009 election for mayor of Burlington, VT, and the 2022 special election for House of Representatives in Alaska. In both cases, there was a clear preference by a majority for a centrist candidate who was eliminated in favor of an extremist.

I think it's a different question for multi-winner elections. I don't have a strong opinion on the best way to run these, largely because it's a much harder question even to determine what makes something the correct result, much less how effectively a voting system chooses that result. The straightforward utilitarian and majoritarian answers don't work here, since they indicate you should just elect as close as possible to several clones of the candidate who would be chosen as the single winner. Intuitively, that's wrong, and we ought to be aiming for something proportional instead, but I don't know how to define that more precisely.