NBest bet would be to have the endpoint only hit the cache, then have some worker which picks up an event to refresh the cache. tRPC won’t do this on its own
NThis is more of an infra question than tRPC. It’s eventual consistency related and one of the harder problems in engineering at scale