We're currently using a setup of ~300MB collections with a 256MB cache size, which means we'll end up in either case 2 or case 3 of Mackert-Lohman. This ticket is to extend the script to include different scenarios, e.g., when the collections fit in the cache.