After a full cache cluster restart, the database is overwhelmed. What is this called and how is it mitigated?