@19
Of course Alekhine was right: objectively and theoretically 3 Be2 is bad. Also Fischer and Kramnik are right: objectively and theoretically 2 f4 is bad. I would not even think of playing it in a correspondence game. However, subjectively and practically 3 Be2 and also 2 f4 may yield good results. It forces the opponent to start thinking as early as move 3 with the clock ticking.
@18
"why Tartakower chose to play that move"
++ He had prepared it as a surprise. He scored 2.5 / 4 with it. It is doubtful if he had scored 2.5/4 against the same opponents playing some main line Ruy Lopez.
https://www.chessgames.com/perl/chessgame?gid=1076244
https://www.chessgames.com/perl/chessgame?gid=1012289
https://www.chessgames.com/perl/chessgame?gid=1076242
Even Fischer played the King's Gambit. Of course not with 3 Nf3, of which he had published the refutation, but rather 3 Bc4. He won, but only after some uneasy moments.
https://www.chessgames.com/perl/chessgame?gid=1044179
A gambit is gambling the position
I get good tactical and attacking position that I like but with the disadvantage of ruining the king safety
It's like giving away defence for stong attack