Just to add something to the third question, there:
A .pnd is its own filesystem. Ponder what happens when you put a CD in your normal computer: The disk is read, and the contents appear as a disk on its own (if you have a windows machine) or appear as its own filesystem, mounted somewhere convenient (on a Linux box). Every single pnd is like a CD on its own, so when you insert an SD card with 10 pnds on, it is as if you had added 10 CDs to the pandora. Small CDs, but still. And the contents of these CDs appear in /mnt/utmp/ as directories (just as a CD would appear under /mnt/cdrom/ or something like that, or under E:, F:, G:, ... on a windows box).
Due to some pnd magic, the content isn't actually appearing until you really try to run an application from within the pnd (you can't insert a SD card and immediately see the contents of all 10 pnds), but the principle is the same. And, just like CDs, you can't write stuff back to these locations (well...Almost). But! The contents of /mnt/utmp/mypnd/ is not on the NAND - In reality, it is inside a file (the pnd) on the SD card.
What do I mean "Almost"? Well, since an application in a pnd might want to store some data persistently, it is given a data directory. When you have a pnd on an SD-card, and the pnd contains an application called, say, "foo", then the first time you start foo, the directory "pandora/appdata/foo/" is created on your SD-card. The application will be able to write data to this directory and read it back later in a way that, to the application, seems as if it was written persistently to the pnd. So: You cannot store stuff inside the pnd, but you can store it in the appdata/foo directory in a way that seems like it is in the pnd.
Disclaimer: I might be confused, lost, and senile. If I am, by all means correct me. However, this is my understanding of how things work.