Good stuff already. But you don't necessarily have to start by testing to destruction. Do some hammock design. Think about your tech stack. What do you think will be the limiting factors under typical user load? The next step is to simulate that load, and address any failures.