- use random DB names in order to detect statements which address a specific DB - split compilation and DB check (DB check and dry run only have to run once) - use separate shell scripts for easier module Travis integration