As early as 1980, computer chess databases were an unknown fænomen, which is why the following could happen.
Two semifinal candidates matches were played in August 1980.
Hubner vs Portich (Italy, Bad Abano Terme)
Korchnoi vs Polugaevsky (Buenos Aires).
Hubner and Portisch played on some August day like this:
and after 61 moves it was draw, but Black missed a win.
The important position is this, after 13 moves:
Hubner played 14. d4 and after ...Bg4, Black gained long lasting advantage.
Bent Larsen was a chess reporter as many times in his life and he analyzed this game and - invited by Naidorf - published it in Argentinian magazine "Clarin" 10 days after this game.
Larsen found better move for White - 14. h3!
Five days after the magazine was published, there was scheduled game Poluagevsky vs Korchnoi.
Korchnoi played the same variation, expecting the same move - 14. d4? Bg4! - just hoping that Polugaevsky couldn't read Spanish.
Soltis (*obviously briefed by CIA) later came with two other explanations, that sound like taken from one of John Le Carre's novel.
1) Polugaevsky's team had - as was a kuytume in Soviet era - had very limited sources, and they wouldn't spend the money for some magazines (they would prefer pizzas or topless night clubs!)
2) The game Poluagevsky vs Korchnoi was played on Sunday. Even if Polugevsky wanted to read magazine "Clarin", and Svoeir embassy might have had it, the embassy was on Sundays closed.
Finally, the game
Polugaevsky vs. Korchnoi