[sherman]

Having never tried with such a large number of books, It's a bit hard to figure out what exactly is going on. I would find KU logs more useful than Calibre logs however.

I wonder if you're running into a memory consumption issue? At startup, KU builds a map/list of all books on device, first by the .metada.calibre file (if it exists), and then by reading all the entries in the Kobo SQLite database, then reconciling the two. This is all done in memory.

It would be interesting to see how much memory KU consumes. If you can SSH/telnet to your Kobo and run 'top' while running KU should give an idea. Sort by memory in top by pressing 'm'. Note, in busybox, it looks like actual memory usage is (%VSZ/100) * VSZ

EDIT: scratch the top idea, doesn't seem to work properly. You can use 'ps -o rss,args | grep /mnt/onboard/.adds/kobo-uncaged/bin/ku' to get a snapshot though.