One of the reasons is that if you momentarily use a very large amount of stack to calculate something (say gigabytes), and then return, the large part of the st

One of the reasons is that if you momentarily use a very large amount of stack to calculate something (say gigabytes), and then return, the large part of the stack which is no longer used stays allocated, full of unused data taking up RAM for the remaining life of the process, which might be a long time. If a heap-allocating algorithm does the same thing, it's likely to free the memory when it's finished with it.