each event is a consumable which means we aren't running large analytical queries to process them.

Our bottleneck is probably infra + optimisations via caching but at the moment I'd say we can handle 1k events per second comfortably :)