By using a doubly linked list, where the first element in the "bucket", contains the next pointer to the next element in the hash table. Read zend_hash.h & zend_hash.c It is fairly complicated and explaining it in depth is beyond the scope of this comment.
This "bucket" also handles the collisions by using separate chaining. There is actually two "next" pointers, one for the chains, and one for the next element in order of insertion. Very confusing and requires reading through the code and playing with it.
You put things in order by memory address and they stay in order when you access them. An array is just a special case of a hash map -- one with a trivial hash function.
That's why the lend themselves to the same syntax so well.