Comment 4 for bug 819971

We could perhaps detect a running instance and skip saving positions for the second one or something, though I'm not sure it's worth the effort.

Ok, I would then trigger saving on configure events, probably delayed or something to reduce the number of disk writes..