When using kernel-side polling, how the liburing and kernel avoid lost wakeup?

Question

After go through the source code,

https://github.com/torvalds/linux/blob/ff1176468d368232b684f75e82563369208bc371/fs/io_uring.c#L6907

I am confuse about the wakeup when use kernel-side polling,

user-thread: A. update-tail, then B. check-need-wakeup-flag

kernel-thread: C. set-wakeup-flag D. check tail

if the user-thread see the order: A. B. C. , and the kernel-thread see the order C. D. A. ，will cause a lost wakeup?