In digging on using logging / reverts on staging I identified some ways in which the data storage definition should be reviewed
- The INNODB storage engine is much better than ARCHIVE - I added the ability to define the table definition by hook
https://issues.civicrm.org/jira/browse/CRM-18193
- The wmf_xx tables are being missed off because they do not start with wmf_
https://issues.civicrm.org/jira/browse/CRM-18180
- The use of the non-unique CONNECTION_ID() as log_conn_id introduces all sorts of complexity that is not helpful