Agree, and added the line to the solution.

That is also my first hunch.

I guess it will turn out to be some very subtle manoeuvring with the kings.
After 1. Kf2, Kd6(b6,b5) 2. Ke3, Kc5 3. f4?, exf4† 4. Kxf4, Kd4 it is a draw.

On the third move White can still try the cheeky 3. Ke2!? and if 3. ..., Kd4? then 4. Kd2, Kc5 5. Ke3 and now it's Black to move in this position which seems a mutual zugzwang. 5. ..., Kd6 6. f4 and White wins.
If 5. ..., Kb4 then not 6. f4?, exf4† 7. Kxf4, Kc3= but 6. d4 and wins.

We're slowly approaching. Black can play better after 3. ..., Ke2...

The study is nice, but the comment is not really enlightening.
The whole point is black defence after
2.Kxa2 Kxc3 3.Ba6 Bxh3 4.Bf1 Bc8 5.Bh3 Bxh3 6.a5 Kb4 7.a6 Kc5 8.a7 Bg2 -+
or do I miss something more obvious ?

Ah nice to see now why you can't take on a2.

The new endgame study shouldn't be so hard with so less material. But i think there might be something i will miss otherwise it wasn't published here.

My first idea is run with the king to e3 and playing f4.

I had seen this endgame study before and I find it still wonderful.

I am wondering about one thing. I do understand that Black needs some pawns on the kingside in order to hold the bishop's ending in case White proceeds more prosaically than in the solution. But is their specific arrangement (f7,f6,g7) of any importance?

Valuable comment, I would say.

Maybe you shouldn't go for f4 but for d4. for the exchange. And 2 Ke2 seems to be better.

For example : 1 Kf2 - Kc5 , 2. Ke2 - Kd4 (or b4) , Kd2 and then you can drive the black king slowly back.