Assess This Ending

MOROZEVICH had a second interesting ending at the 2001 super-tournament in the Kazakh capital Astana.

White to make his 41st move in Morozevich (2749) - Kasparov (2827)

White is temporarily a pawn up but Black has the more active king and bishop. What should the result be - a definite win for White; White is better but not necessarily winning; draw; Black is better but not necessarily winning; a definite win for Black?
*****
*****
*****
*****
*****
"All White's pawns are horrible and so his position should be lost," write IM Vladimir Barsky and GM Alexander Baburin in the October 2001 edition of Chess Life.
41.Bc4 Bxe4
In ChessBase's 2019 Mega database, GM Ľubomír Ftáčnik (or is it Luc Winants? Both are separately credited as annotating the game) comments: "Despite equality in the number of pawns, (the) endgame is very difficult for White (as) Black enjoys (the) better pawn-structure and more-active king."
42.Re2 f5 43.Bd3 Kf4 44.Bxe4 fxe4 45.Rf2+ Ke3 46.Rf8 e5
Barsky & Baburin give this an exclamation mark but (with the help of Komodo9 and Stockfish10) it is far from clear that the text is any better than getting on with it with 46...Ke2.
47.Re8 Rc7 48.Rxe5
Barsky & Baburin give this a question mark, saying White should have activated his king with 48.Kg3. The analysis engines disagree.
48...Kf4
This gets an exclamation mark from Barsky & Baburin. Stockfish10 agrees immediately that it is the best move, and Komodo9 follows suit if given enough time. However, the engines agree 48...Rxc2+ also wins easily.
49.Rb5 Rxc2+ 50.Kg1 e3 51.Rxb4+ Kf3 52.Rb1 Rg2+ 53.Kh1
If 53.Kf1 then 53...e2+ 54.Ke1 Rg1+.
53...e2 54.a4
The sort of despairing move I have played many times - it clearly does not achieve anything, but what else can White do (apart from resign)?
54...Kf2 55.a5 Rg5 56.Kh2 Rxa5 57.h4 Ra3 0-1
Getting back to the question I posed at the start, the engines (as well as Barsky and Baburin) reckon the diagrammed position is definitely won for Black.