That's fair but it's trivial to convert a BinaryTree to a LinkedList (doubly or singly). If you eliminate the requirement that this must be done in-place (which implies O(1) auxilliary space), then the solution is not particularly interesting.

The practical use case for in-place algorithms is operating on large data sets that are directly on a disk. Think some kind of database organizing data directly on a hard drive. You want to avoid having to double the number of hard drives you need in order to do some kind of data transformation, and perform any linking/relinking directly on the disk.