Top player meet fairly often, but about half of their matches end one draws. A simple model of 25% win, 50% draw, 25% lose gives a top player a 1/4^7 change (about 1:16000) of winning seven in a row against another top player.
Maybe not, but I bet one can model the results of chess matches as some random noise over the expected probability distribution of "player P against player Q for every pair of players (P,Q)" fairly well.
It is surprisingly hard to show that any sports time series is not random. 'Streaks' in baseball, for example, do not really exist according to some statisticians.
http://www.64to1.com/2014/09/04/expected-wins-in-top-level-c... has more info, but, surprisingly, doesn't do this simple calculation, which I think is a better model.
So, you would have to play about 2500 of these tournaments to get one such result where one player wins the first seven games.
Cases where one player scores seven points or more in ten games are way less rare, as there are 10 over 7 or 120 ways to get seven wins in ten games.