⁵According to the professor: When a query is received, the cache is visited to see if there is a matching entry. If it is found, it is used to generate a plan for the current one, which will be then tested for some time. If the generated plan takes too long or there is not any query pattern in the cache matching the current query, the system launches different executions in parallel using alternative access paths. Eventually, one of such executions will finish and the others will be killed. The executed one will be the one kept in the cache.