Double checking if there's anyone out there that can confirm this.
After reading this again, I am leaning towards the entire shard being cleared, not individual documents.
The requests cache is a very good fit for the logging use case, where only the most recent index is being actively updated — results from older indices will be served directly from the cache.
For the logging case, seems like what's mentioned here is that logging to new indexes or new shards would mean old ones don't get refreshed, and don't have their caches invalidated. Guess there's no way to do that on a shard that is being updated, even if it's just appends.