“Rosetta 2 translates the entire text segment of the binary from x86 to ARM up-front.”
The software is clever, but the cheat code is in the silicon. Apple put x86-style memory ordering in the M1, plus an undocumented extension to compute x86 parity and adjust flags. Everyone else’s emulator has to fake those in software. Owning the chip and the OS lets you build the translation layer into the hardware. Nobody else gets to play that way.