I like the point about not being able to trap the bishop as I didn't even consider that in my analysis of the position. A few GM rated players thought that Bxd5 was actually forced at first and got h6 from a computer from what I've read so I think this oversight is only easy after you have been given the answer. After knowing the sequence it is very easy to overestimate the simplicity of the position.
My type of position studying
I really like this position. While the move might be pretty easy to find, perhaps the reasoning behind it is a bit more elusive for a player of my level.
In this position, the move that I would probably play would be 1.Rcd1 most likely. Just to double my rooks since I'm expecting 1..Rbb2 from black. 1.a3 makes a lot of sense but I think that would be giving myself too much credit. I would definitely like to think that I would play the latter.
I would not of been thinking that black's plan is to play 1..Bxh3 with a basic knight and rook checkmate pattern resulting if you try to take the bishop with the pawn [2.gxh3 Nf3 3.Rd2 Rb2 34.Rd7+ Kh6 35.Rxh7+ Kxh7 36.Rd7+ Kh6 37.Rh7+ Kxh7 38.Bg8+ Kxg8 39.Ne6 Rh2# ]. Without taking the bishop, black has a much stronger attack now after something like 1.Rcd1 Bxh3
Given black's plan above, the only 2 moves that come to mind are 1.a3 and 1.Nd3 which I think 1.a3 is just more natural and it turns out that it's better anyway.
Maybe I just like this position because it is a reminder of how important pattern recognition is. This position would be particularly fun to play as black though.
All in all, this position is pretty neat to me.
Magnus Carlsen played that move. It is said that after 1..Bxd5 2.Bxg7+ Kh7 he missed the move Bd4. So, naturally I saw that it attacks the knight on a7 but so what. Can't I just move the knight to c6? Then it all comes together in your calculations that 3.Bd4 Nc6 leads to 4.Nh6#. All this has to be a part of your calculations from the above position. You have to see that if the knight on the f-file moves the white rook controls that file completely now. You have to see that 1..Bxd5 leads to you giving up your rook for a bishop. You have to see that you should play 1..h6 instead.
This is a very daunting task for someone at my skill level who has no chance of seeing this without having been exposed to something similar and even a slim chance then. I would like to see positions like this where a GM missed a certain move during they're analysis because, if they are anything like this one, I find this very instructive when it comes to calculations. This one is recent and caught my eye.