I've looked at the code, it appears that if innodb_import_table_from_xtrabackup is enabled in my.cnf, InnoDB will scan _all_ tablespaces on startup, even those that have no corresponding .exp file. Which is of course unacceptable in production. So yes, we have a problem here.
Since in case of a crash on table import, restarting the server with improved import disabled is impossible, I think we should fix this by making it possible to start the server with innodb_import_table_from_xtrabackup enabled. That is, only scan tablespaces for which an .exp file exists, and remove it on successful import?
I've looked at the code, it appears that if innodb_ import_ table_from_ xtrabackup is enabled in my.cnf, InnoDB will scan _all_ tablespaces on startup, even those that have no corresponding .exp file. Which is of course unacceptable in production. So yes, we have a problem here.
Since in case of a crash on table import, restarting the server with improved import disabled is impossible, I think we should fix this by making it possible to start the server with innodb_ import_ table_from_ xtrabackup enabled. That is, only scan tablespaces for which an .exp file exists, and remove it on successful import?