Hi, In a very busy library the real time hold queue can get significantly behind (1500-1600 entries, for example) when many returns are processed at once. We're just wondering if there is some inefficiency there that we could resolve to make things move through more quickly (batching of requests was my first thought). In fairness, currently this is using database polling and not rabbitmq - the answer very well could be "use rabbit and this will improve" - which actually is an answer I'm willing to accept, if not immediately implement again in this way.