AylerKupp: <<saffuna> Once it's down to the bishop endgame, how much would be analysis and how much is technique studied and mastered by GMs?>
If by analysis you mean OTB calculation then I think that you'll agree that there is less of an opportunity to do deep calculations/analysis at fast (rapid, blitz) time controls than at classic time controls, unless one or both players at the classic time control are in time trouble. Therefore, usually, I would think that the top players would rely more on technique and less on OTB analysis/calculations at the faster time controls than at the classic time control.
Technique is the application of a skill developed through study and experience, plus learning and memorization (e.g. achieving a Lucena position). So I would say that in this and other endgames at faster time controls the top players rely more on technique than calculation; the players simply don't have enough time to do a complete analysis of complicated endgames.
Having said that, it's obvious even to me that after 19...a4 White Pb5 and Pb2 are not easily defended given Black's control of the light squares around them. So Black's winning "technique" would first start with winning that pawn, and for Black to this (a) either Black's bishop must be placed in the a4-e8 diagonal or his rook on a5, (b) White's rook must be prevented from establishing an outpost on e5 defending the Pb5, and (c) White's Nd4 must be chased away so it can't defend the Pb5. Then it is a question of the calculations needed to make that happen, something that So accomplishes in moves 20 – 34, after which the Pb5 is doomed.
After that the issue is whether a single pawn advantage in this BOC endgame is winnable for Black. The knight exchange, usually not desired when you're down in material, is practically forced since otherwise White's bishop is tied to the defense of the Pb2 forevermore. Then the technique for Black is just a question of marching his king to the q-side to attack, or threaten to attack, the Pb2. To do that, the Black pawns must be placed on light squares where they cannot be attacked by White's DSB, and Black's bishop must keep White's king away from the q-side and hold back White's k-side pawn majority, something that So accomplishes by 46...Bg6+.
Black can then get a 3:1 q-side majority by force even if it means giving up his Pe5. After that it's only a question of advancing the a-side pawns until he obtains 2 passed pawns, and So accomplishes this by 53...c3+. Note that Black would win even if White had marched his king to g4 since Black could give up his bishop after h4-h5 ...(any) hxg6 ...hxg6 since it would take White a long time to capture Black's Pg6 and long before that White would have had to give up his bishop to prevent one of Black's pawns from queening.
So I would say that this endgame is almost all technique and the only analysis/calculation needed would have been to avoid a gross blunder such as giving up his bishop without compensation. Of course, this is just one example, but I would think that, in general, at the faster time controls top GMs rely much more on technique than analysis/calculations. Then again, I'm not a top level GM so I wouldn't really know. Best to ask So directly; send him an email with your question and see if he responds.