Our database was overloaded by too many concurrent device registration requests. As a temporary measure, we upgraded database capacity and implemented a simple registration queue. To prevent future incidents, we will implement a more robust rate-limited queue to handle sudden spikes in registration.