Does a relaxed memory-order spinlock always break synchronization?

Question

Consider the following code:

int nonatom = 0;
std::atomic atom{0};

// thread 1
nonatom = 1;
atom.store(1, std::memory_order_release);

// thread 2