Let the servers keep crashing, tell everyone to add new instances to help with performance, which puts 1500 rows into the database tables that used to have 50 rows and invokes a massive federation 1-vote-1-https overhead… causing more crashing… all the while ignoring the SQL design of machine-generated ORM statements and counting logic hidden in the background triggers.
… keep users off your sever as a method of scaling by crashing. It’s one of the more interesting experiences I’ve had this year! And I spent all of February and March with the release of GPT-4… which was also interesting!