https://jira.nypl.org/browse/SIMPLY-2015 represents a big improvement, cutting 25% off of our grouped feed generation time, but that's not enough for what we need. Grouped feed generation still needs to take place in the background.

Fortunately, while investigating a follow-up I was able to cut an additional 75% off of our grouped feed generation time by fixing some obvious mistakes and applying a few simple optimizations.

These optimizations take basically all OPDS feed generation into the realm of "can be done while the user is waiting for an HTTP request" – a huge positive step.