I imported an XML file with 1760 records just now, and the file imported perfectly. BUT, that was my 10th attempt at least. I had to fiddle with the file many times before I got the import to work properly. Import, check, delete the error records, try again...
This is not the fault of the importer: the data file must be groomed first. I compared my file with a sample export file from the BookLibrary system. And made many corrections using search and replace.
This is always true when doing CSV or XML imports, it is very fussy work. Now that I know what the importer needs I will be able to prep my other files properly the first time around.
So, I would say that the import feature works very well, in my opinion. To import 1760 records perfectly is a nice accomplishment.
My own question: whether the system can manage a library with 100,000 to 200,000 records... apparently I will have to try it and then I'll know.