The API docs state the battle logs are cleaned out every 7 days.
The battle logs on my capitol current scroll back to 26/01/2012 - almost a year.
I changed the 'delete_all' (which deletes records one-by-one) to 'delete' (which deletes them via the database)
This is safe to do in this instance since there are no method modifiers on the delete method.
That works, but I will see if, now that the bulk of the messages have been deleted, it works as designed.
As a side, we perhaps need a flag/email/report (other than the log files) when a script fails to finish correctly.