This is an interesting problem. 

I think your suggestion

> My best guess is that doing an st_unmount(st, NOEJECT) in stclose()
> when the tape's at BOT would be an improvement over the current
> situation.

is a good one. I'd say either do it or send-pr it.

What's odd is that the PREVENT condition is usually overridden by tape
changer actions- but I guess it depends upon the changer. Some changers
have extra input signals into drives so that the PREVENT condition can be
overridden (e.g., LAGO DataWheels)- but I guess from your description this
one doesn't.


You have another real problem, which is the automatic loading of the next
tape in the changer. Mostly that's a strappable option for changers- you
might look into changing that so that you don't get surprised by what gets
overwritten.

-matt