how does TOC iterate over time?

[greenlight2000]

from what I understand from the paper (especially the prompt demonstration), RAC's input is the root node ambiguous question(AQ) from the test set, (plus some retrieved passage), and its output is a series of disambiguated question(DQ) with its answer. And that's it. How does the TOC execute RAC iteratively over time? does it turn one of the generated DQ into a AQ and do further RAC on it? It seems obscure to me and the whole paper didn't put much explanation to it. Can you please clarify it for me? I cannot picture an AQ being clarified though multiple round of RAC, would you provide an example of it?

eg. how does this work? “What country has the most medals in Olympic history?”(AQ) -> "What country has the most total medals in Olympic history?"(DQ) -> “What country has the most gold medals in Olympic history?”(DQ of DQ?)