Could implement Redis caching or similar to keep track of failures across different receiver instances. We'd want to log the final failure (aborting) into a DB or something.