If you read back through this thread you would understand why it is different behaviour between running one versus multiple. Calibre has a major issue with memory leaks in the conversion process, so to work around this conversions should be done in the background. However if you are just doing a single ad hoc extract ISBN (which is how I usually tend to work) then for speed reasons I don't run it as a background job if you select only a single book.
It sounds like Calibre is crashing on your books when doing the extract when running in the background. No-one else has reported any issues with this, so I am inclined to believe it is something about the books you are scanning. You need to figure out the format that is causing the issue - duplicate the books (create empty books then merge in keeping the original), then one by one remove likely problem formats (starting with PDF) to see if it still errors.