The Compute Engine step "Purge db" can be optimised by dropping a few operations:
- do not delete old records of ce_activity and ce_scanner_context because they are already purged by the "startup leader" server on each new deployment
- do not try to delete past measures of directories and files. Indeed the table project_measures contains only project measures.
- do not persist the measures on new code in project_measures. They are never read and past measures require to be deleted on each new analysis. That generates a lot of IO in DB.