I have not watched the video, so I'll go by your description. There is an array of items which contain pointers to other items. These pointers form a linked list structure.
This sort of data structure tends to arise when a collection of items needs to be easy to traverse in multiple different orders. For example, a priority queue may need to be traversed both according to the elements' age and according to the elements' priority. This can be implemented with an array of elements, with new elements added at the end so that they're in age order, and a linked list or a search tree for priorities.
One common case for linked lists of array items is in a memory allocator. You ask, “Why use these free nodes, when you can dynamically allocate storage?” Well, what about the cases when you can't dynamically allocate storage, because you're implementing the dynamic storage allocator? The concept of “free list” is common to most memory allocators. A block of memory that isn't in use is an element of the free list. With fixed-size blocks, to free a block, you write a pointer to the current head of the free list at the beginning of the block, and you set the head of the free list to the address of this block. To allocate a block, you take the current head of the free list as the block to use, and set the head of the free list to the block that
was second in the list. In pseudocode:
block := free_list_head
if free_list_head is null: raise out_of_memory
free_list_head := free_list_head.next
block.next := free_list_head
free_list_head := block
Allocators for variable-sized blocks typically use fancier data structures so as to be able to merge adjacent blocks and find blocks of an appropriate size.