When Stockfish 9 came out, I did a speed chess round robin between Stockfish versions 6, 7, 8, and 9, and all the games were draws except for the games played by Stockfish 9, which won all its games against the older versions of the engines. That is a significant difference.
I analyzed one position with Stockfish 9 and the new version 10 running at the same time. After 5 minutes, Stockfish 10 was thinking 1 ply deeper in the same position.. Is that significant? Probably. For most positions the engines would be very similar, but being able to think half a move deeper than your opponent usually makes you unbeatable, because you already see all the variations that your opponent is going to think about.
Both programs ran at roughly the same number of nodes per second. Given that this is a 27-28 ply search, achieving 1 ply deeper probably means that 10 has to eliminate about 3% of the variations that 9 looks at, by my calculation. This could also mean that 10 has a weakness that 9 doesn't. Or this could mean that 10 has a better evaluation function, leading to more variations being eliminated.
There were some higher rated players who dropped out before the final round, probably because they lost to the top player with the monster rating. It says "17 players", but that doesn't include the 18 who dropped out before the final round.
I had a couple of games that I really should have lost, but won anyway. I made many errors, but so did my opponents. This happens sometimes in speed chess. My last round opponent literally had me in a mating net and threw it away.