Gilmoy: I watched that MatoJelic video, but ... he didn't <explain> anything after <20..Ne8>, he just recapped the game line only.
White clearly has the <Morphy mate> on g: <discovered check controlling h8, with f8 self-blocked>. Since this mate requires only R+B, White can offer the other 3 pieces -- and does offer 2 of them. (A true Morphy would actually make the opponent accept all 3, tsk --)
1. Why <24.a3>? It must be a deflection to lure Black's Q away from e7. From e7, Black would have a desperado doubled interposition Qg5, which crimps the Qg4+ lines. I think White still has a won endgame on pawns up, but it's indeed inelegant.
2. Why <27.Nxd5>? I suppose Nf6 could spite-stop the Morphy mate pattern in two ways: Ne4 attacking the Rg3 (so the discovered check doesn't work), or Nh5 defending g7 for one more tempo. Ergo, it's overworked: it can't keep those options alive <and> defend d5. And White doesn't need the Nc3 anyways, so it might as well smite for mayhem. Oh-by-the-way, it does double on Nf6, which now has not enough defenders (see <24.a3>, and nod). As soon as Nf6 abandons its post <27..Nxd5>, the Morphy mate (and the Rg8-etc-Qg4+ subtree) becomes unstoppable.
Ne5 quietly grips f7, which snuffs all f6/f5 defensive tries. Normally we sac or blockade on f6 to prevent f7 from moving, but here it doesn't matter. Similarly, h6 isn't luft because Qf5 still owns h7.
<17..Rf8 23..Be8> were ultimately too defensive, ceding a 5-v-0 build-up in the center. Overprotecting f7 suggests (invites? dares?) that White should sac at not-f7. Probably <18..Qb4> triggered a deep-think, and White saw the b2-as-decoy and g7-as-target themes, hence he worked out <20.Rd1!> offering b2 while enabling a second Rook lift.
Black needed more <willing-to>. Sally forth, vie for the board, do some building-up of his own, get off the sofa along 8. It's a rare win for White because Black doesn't normally allow these things to just happen.