My bad, replication was stopped at the time so there were no queries or transactions occurring during the backups as it's just a backup slave, which meant that there was nothing to replay or recover. Doing a backup at the moment with replication running and it's gone back to doing it's usual thing of taking quite a while. I'm getting the "1 transaction(s) which must be rolled back or cleaned up" message again which was a pre-cursor to the assertion failure before, so hopefully this will succeed.
My bad, replication was stopped at the time so there were no queries or transactions occurring during the backups as it's just a backup slave, which meant that there was nothing to replay or recover. Doing a backup at the moment with replication running and it's gone back to doing it's usual thing of taking quite a while. I'm getting the "1 transaction(s) which must be rolled back or cleaned up" message again which was a pre-cursor to the assertion failure before, so hopefully this will succeed.