Nerwal: <I thought it was a book draw with proper technique.>Many positions with this material are drawn (but far from all; in Keres vs Sokolsky, 1947 white is already winning after 63... ♖a1, with the pawns still on f3 and h4!), true, but in practical play white's winning chances are still quite high. Even elite players have lost it (quite recent examples : Grischuk vs Eljanov, 2008 Kramnik vs Aronian, 2008 ).
Muminova started to go wrong as early as move 64 :
64. ♔f2 loses while 64. ♔g2 draws. But the winning manoeuver after 64. ♔f2 is extremely complex, starting with 64... ♖b3+ ▢ 65. ♔f3 h3 ▢ and the pawn goes to h2.
64... ♔g4? only draws.
65. ♖a4+!, only drawing move.
66. ♖c4 is losing to 66... ♖b2+.
73. ♔e2 is the wrong side (73. Kg2!)
75... ♔g3 only draws. Playing the rook on the second rank wins; this is the main difficulty of this endgame, knowing when to transpose into a winning ♖+♙ vs ♖ endgame, although here it's easy to figure that ♖xh4+ ♔g3 mate or winning the rook loses.
78. ♔f1 is the decisive mistake; black should wait along the g file. It's not easy to see why; one explanation is that the rook doesn't have the checking distance after 81. ♖f5 ♔g3 82. ♖g5+ ♔f4; this defence would draw were the rook on last rank.
After 78. ♔f1 Anna Muzychuk played perfectly to convert. Muminova played on a bit too long, maybe hoping for something like 92... ♔f4 93. ♖h8 ♖g1+ 94. ♔f2 h1=♕ 95. ♖xh1 ♖xh1 stalemate.