DOUBT on INACURACY
Move 31 correctly identifies the d6 pawn as a weakness since it looks backwards (although more accurately it is isolated and blockaded by the d5 knight), but c5 was a mistake because it gives black the chance to trade away their weak pawn.
Move 32 was an inaccuracy according to the quick analysis probably because Nf4 was the slightly better move, but 32. Qxc5 was fine in my opinion too. Nf4 forks the queen and e6 bishop, so it provokes black into playing ...Bxf4 and this makes black give up the "bishop pair" which is often evaluated at roughly half a pawn of material.
Move 33 was innaccurate for the same Nf4 move being considered slightly better by the post-game analysis for the same reason.