I just finished this game.  It was a correspondence game from the thematic team match between “French Defence fanatics” and “Team Belgium” on chess.com, I was playing on the 1st board for “French … ” team.  I lost pretty badly the first game with Black and wanted revenge. I played also 2 games with this guy more than a year ago, 2 draws.
It started with the variation that Korchnoi and Karpov played in Candidates match in 1974, 3…c5 in Tarrasch leading to Black having isolated pawn on d5. Their games were positional, here I castled queenside and it became pretty sharp, with White pawn storming on the kingside.
My opponent for some reason didn’t respond with the same and restricted himself to defense. I sacrificed a pawn to open the lines. Then my opponent missed the moment, when he could equalize (after 25. h6) and after a few mistakes got essentially worse.
I thought my 35. Bf6 was good – winning an exchange, but Fritz says I had better move 35. Rd8+ winning a piece and a game. I think I saw it, but I believed that the king escapes after 35… Kf7, not seeing 36. Qh5+ winning then a queen for a rook.
Anyway I went up an exchange and in occurred R vs. N endgame had to be careful with him having an “h” passed pawn. After 47. Rg1+ I think he got scared of the mate ( my king was on e4 ) and abandoned the “h” pawn, taking instead my pawn on a2. After a few more moves my advantage became decisive and he resigned.

I recently played a blitz game, where for the first time ever I got this ending. I had pretty remote idea how to win, pursued his knight (ignoring the draw offer) for 48 moves until he blundered.  Of course I wasn’t quite satisfied, so I decided to learn this ending.

Dvoretsky says that it is usually drawn, but there are several exceptions:
1. If the knight separated from the king, often it can be caught.
2. If the knight is in the corner, it can be taken because of zugzwang.
3. The squares g7/b7 are not good for the knight.
So, I looked at my game and found the moment (mostly my opponent played right),
where I could separate and win his knight. It is not easy and even Fritz had a problem,
so I used Endgame Nalimov Tablebases Online. Here is the result:
What I also learned that there are typical ways of attacking/restricting the knight,
for example putting the rook nearby on the same vertical/horisontal.
Though this ending is rare, I found it useful for learning about
relative strength/interacting of the rook and knight, which should be useful
for the regular endgame (with pawns).

Dvoretsky says that it is usually drawn, but there are several exceptions:

1. If the knight separated from the king, often it can be caught.

2. If the knight is in the corner, it can be taken because of zugzwang.

3. The squares g7/b7 are not good for the knight.

So, I looked at my game and found the moment (mostly my opponent played right), when I could separate his knight and king.  Winning is not easy and even Fritz had a problem,  so I used Endgame Nalimov Tablebases Online.  Here is the position after Nf5-e3?:

RN1

1. Rf7 Ke8 2. Rf4 Kd7 3. Kd4

RN2

3. … Nc2+  4. Kd5 Ne3+ 5. Kc5 Kc7 6. Rf7+

RN3

6. … Kd8 7. Kd4 Nc2+ 8. Ke5 Na3 9. Rb7

RN4

9. … Nc4+ 10. Ke6 Kc8 11. Rb5 Ne3 12. Rb2

RN5

12. … Ng4 13. Rc2+ Kb7 14. Re2 Ka6 15. Kf7

RN6

15. … Nh6+ 16. Kg6 Ng8 17. Re8 Nh6

RN7

Though this ending is rare, it shows very well the relative strength of the rook and knight.  Also you can learn the typical ways of attacking/restricting the knight, for example putting the rook nearby on the same vertical/horisontal which should be useful for the regular R vs. N endgame  (with pawns).