We give you a service that works fine on your laptop and falls over at four hundred requests a second. Your job is to find out why, six times.
Each block introduces one failure: a connection pool that never drains, a retry storm, a cache that stampedes, a queue with no dead letter. You reproduce it, you measure it, you fix it, and you write the postmortem.
By the end you will have a service that holds, and — more usefully — a method for finding out why one does not.
