The trick with simply replaying a db-api log for regression testing is that you don't necessarily end up with the same primary keys if you rebuild an openacs instance from scratch. Once these get out of sync, your log is only useful if you backup/restore the initial db (and then you're not regression testing the installer script).
I suspect that capturing the web server log and replaying that through tclwebtest or portable.nsd might be a better idea.
I've had good luck with the latter... it's robust to user interface changes and it lets you test at a more granular command vs page level. YMMV.
Try a google search on "Unit Testing Database code" for more info on the issues involved.
