Why allow `propagate_on_container_swap == false` in Allocators, when it might cause undefined behaviour?

Question

^{Note: Originally asked by Matt Mcnabb as a comment on Why can swapping standard library containers be problematic in C++11 (involving allocators)?.}

Answer 1

It is not so much that the Standard allows propagate_on_container_swap to cause Undefined Behavior, but that the Standard exposes Undefined Behavior via this value!

---

A simple example is to consider a scoped allocator, which allocates memory from a local pool, and which said pool is deleted when the allocator goes out of scope:

template 
class scoped_allocator;

Now, let us use it:

int main() {
    using scoped = scoped_allocator;

    scoped outer_alloc;
    std::vector outer{outer_alloc};

    outer.push_back(3);

    {
        scoped inner_alloc;
        std::vector inner{inner_alloc};

        inner.push_back(5);

        swap(outer, inner); // Undefined Behavior: loading...
    }

    // inner_allocator is dead, but "outer" refers to its memory
}