If the child finishes before the parent, the kernel must keep a small amount of information available in case the parent later calls waitpid.

The kernel can reclaim other resources such as memory and can close any open files, etc. and the child process will never be scheduled to run again.

Once the parent calls waitpid, the child process is said to be reaped by its parent and the child process can be completely deleted.

If a child terminates before the parent waits for it, this unwaited for, but terminated children hangs around and is called a zombie processes.

If the parent exits without waiting for one or more of its children, the kernel can clean up the zombie children (but not children that haven't yet terminated.)