We've restored service to the queueing system and predictions are flowing again.

We've identified what's gone wrong (an internal queueing system is experiencing a disk error) and we're working on restoring service.

We're investigating issues with prediction creation.