In the 8th game of the series1 between AlphaGo Zero and AlphaGo Lee the following position with a double ko at {A,C}×{13,14} arises: On the next move AlphaGo Lee (Black) plays the double ko exchange A14-C13 before its next effective move at T15 and subsequent three effective moves; then AlphaGo Zero joins in the fun and also makes double ko exchanges before its effective moves. This continues until move 272; after 276 Black resigns.
I do not see the benefit in these exchanges as against immediately playing at T15 etc. Are these exchanges useful or merely something that their respective training never taught them how to do? If they are useful, what purpose do they serve?
1 See the article in Nature at https://www.nature.com/articles/nature24270#t1; the full text is not free but selected game records are, in the download at https://media.nature.com/original/nature-assets/nature/journal/v550/n7676/extref/nature24270-s2.zip