QEMU Translation block

Question

I'm studying how QEMU translates native basic blocks and runs them. As far as I understand, it reads native basic blocks then retranslates them, but I wonder how it can deal with exceptions inside, so I run the following program:

// pgfault.c
int main() {
    __asm__(
        "xor rax, rax\n\t"
        "mov rbx, [rax]\n\t"
        "xor rcx, rbx\n\t"
        "add rax, rcx\n\t"
        );
}

// compile
x86_64-w64-mingw32-gcc pgfault.c -masm=intel -o pgfault.exe

// and run in PANDA/QEMU

Indeed, I get the trace:

// first translation block
0x401510           xor rax, rax    ; first instruction of main
; second instruction generates a page fault

// second translation block (KiPageFault in ntoskrnl.exe)
0xfffff80026e4600  push rbx
0xfffff80026e4601  sub rbp, 0x158

The second instructions generates a page fault, then the translation block contains only the first instruction. How can QEMU/PANDA knows that ahead of time (i.e. at translation step)?