I assumed, at first, that it was somehow falling through the infinite loop and accidentally runnning the unreachable function, but it clearly explicitly runs it in the assembler generated...
10f4: 48 8d 3d d5 00 00 00 lea 0xd5(%rip),%rdi # 11d0 <_Z11unreachablev>
10fb: ff 15 b7 2e 00 00 call *0x2eb7(%rip) # 3fb8 <__libc_start_main@GLIBC_2.34>
how odd.
edit: ah, it's called from __start, which suggests that main is being elided entirely by the optimiser, and somehow 'unreachable' is simply becoming a defacto 'main'