Here's what my sources say about this endgame.
Müller, Lamprecht, Fundamental Chess Endings:
With three against three on one wing the defender is usually able to hold on.
[Title "Mueller vs Yusupov, Hamburg SKA (1991), after 66... Ba5"]
[fen "8/5p2/6p1/b1n4p/2k2P1P/4K1P1/1R6/8 w - - 0 0"]
1. f5 Bc3 2. Rg2 gxf5 3. Kf4 Kd5 4. Kxf5 Ne4 5. g4 Nd6+ 6. Kf4 Be5+ 7. Kf3 hxg4+ 8. Rxg4 Ke6 9. h5 Nf5 10. Rg1 Kf6 11. Rg8 Nh6 12. Rg1 Kf5 13. Rg2 Bf4 14. Rg1 Bg5 15. Rg2 Bf4 16. Rg1 f6 17. Rg7 Bg5 18. Kg3 Ng4 19. Rg8 Ne5 20. Rg7 Bh6 21. Rg8 Nd3 22. Kh4 Nf4 23. Rh8 Bg5+ 24. Kg3 Ne6 25. Rg8 Bh6 26. Rg6 Bf4+ 27. Kf2 Bg5 28. Rg8 Nf4 29. Rh8 Kg4 30. h6 Kh5 31. h7 Kh6 32. Kf3 Ng6 33. Ra8 Kxh7 34. Kg4 Ne7 35. Kh5 Kg7 36. Ra5 Bf4 37. Ra6 Be5 38. Rb6 Kf7 39. Ra6 Nc8 40. Rc6 Nd6 41. Ra6 Ne4 42. Kg4 Kg6 43. Rc6 Bd6 44. Ra6 f5+ 45. Kf3 Kf6 46. Ke3 Kg5 47. Kf3 Nd2+ 48. Ke2 Bf4 49. Ra4 Kg4
White should just sit and wait in his fortress. However, he decided to start some misguided activity.
Dvoretsky, Dvoretsky's Endgame Manual:
If both sides have three pawns placed on the same wing, then the defender may hope for a draw (only if, of course, the pawn structure is devoid of grave flaws).
[He then analyses the endgames in Capablanca vs Lasker, St Petersburg (1914) and Timman vs Karpov, Bugojno (1980).]
Flear, Practical Endgame Play - beyond the basics:
The case where all pawns are on the same flank is interesting as the result isn't always evident. However, a general rule could be that the bishop and knight require a target in order to convert their superiority.
[He then uses Ponomariov vs Sasikiran, Biel (2004) and the aforementioned Timman vs Karpov, Bugojno (1980) as examples.]