Interesting. They are implementing, in effect, an LRU cache algorithm: scanning books as they are requested. The next person to request that book gets an email in 30 seconds, not two days.
What are they discarding? It's more of a FIFO isn't it? I assume that when they have no new requests they are scanning books based on some other criteria.
Perhaps it's best described just as an infinite-sized cache. I would assume they have enough request to very rarely be idle.
Copy-on-write might be a better concept to allude to. You're making a new version, but efficiently, on demand.