(And I still say B+Kt is hard...)
Hardest endgame to master?

The hardest one for me was always (and continues to be) N v B.
That's because it's unwinnable!
I think he meant N vs. B with pawns.
Personally, I've found queen endgames in general to be the hardest because they tend to be the most demanding on your ability to accurately calculate, and theoretically won positions can be difficult to actually win in practice. At the very least, many of them require exorbitant patience as you watch your opponent check, check, check you, hoping that you'll eventually be able to take over the initiative yourself and make progress.
I've read that BB vs. N (no pawns) has been shown to be a forced win for the side with the two bishops (assuming they don't operate on the same color diagonals) via computer analysis. However, I think the win can be up to 278 moves long or something, so in practice, it's probably safe to say it's a draw.
Perhaps the hardest endgame to win in a game I've played is bishop + wrong-colored rook pawn vs rook pawn. In fact I offered a draw because I didn't see a way to win. (It was a theoretical mate in 55.)

Defending the worse side of R vs R+N for a draw seems hardest to me
Defending the worse side of R vs. R + B is even harder. I wonder how many super GMs successfully defended theoretically drawn R vs. R + B endgames or reached move 50 without a clear mate in sight in theoretically lost R vs. R + B endgames. I definitely agree with all the other endgames mentioned here, but defending R vs. R + B is crazy.
my top 10 pawnless endgames (not that accurate because i don't know much about endgames)
10. King and Queen vs. King
9. King and Rook vs. King
8. King and 2 Bishops vs. King
7. King and Queen vs. King and Bishop
6. King and Queen vs. King and Knight
5. King, Bishop and Knight vs. King
4. King and Queen vs. King, Bishop and Knight
3. King and Queen vs. King and Rook
2. King and 2 Bishops vs. King and Knight
1. King, Rook and Bishop vs. King and 2 Knights
The really hard endgames are in the category of 2K + 3 or more other units. For instance K+2N vs K+P, K+2B vs K+N, K+Q vs K+R+P, K+2N vs K+R+B. Most of these endgames were not completely mined until the tablebases of the 21st century were constructed . Commonly they are too hard to master for a human which is why they are absent in endgame courses.

I can't remember where I saw it but I can recall having seen that Q vs BN+rook pawn is winning for the queen side.
This link has a training dedicated to it, but it didn't say that this is a theoretical winning endgame: https://www.chessvideos.tv/endgame-training/queen-vs-bishop-knight-h-pawn-13.php
Seems hard enough to me.
Edmar Mednis has estimated that 4% of R+B vs R starting positions in games are theoretical wins...
lol