Related question: why is the stack limit so tiny in modern operating systems? Why do we allow programs free reign of the heap but not the stack?
Edit: I hadn’t considered that all of a processes’s threads share the same address space. In a 32-bit address space, a 4MB stack limits you to 1024 threads (with nothing left over for heap).
However, perhaps the question still stands: why use a small stack size on a 64-bit system.
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.
Edit: I hadn’t considered that all of a processes’s threads share the same address space. In a 32-bit address space, a 4MB stack limits you to 1024 threads (with nothing left over for heap).
However, perhaps the question still stands: why use a small stack size on a 64-bit system.