Audittrail collection becomes slow with many audit trails

To figure out which is the next audit trail sequence number is the next to be collected for a given contributor in a given collection, the service needs to join two (large) tables and perform a sequence scan.
This works fairly ok, when the service have few audit trails, but trashes the database with diskreads when trying to figure out how which is the start of the next page.
A simple solution is a table for keeping track of what have been collectected, i.e. a table containing:
Collection | Contributor | latest sequence number