A reference library of 10,000 flows has the same enemy as any aggregated dataset: the products keep shipping after you capture them. We keep a few thousand job listings from dozens of sources current, and adding new entries turned out to be the easy half. The hard half is noticing that something you already hold has changed or quietly disappeared at the source, because nothing announces it. For a product team moodboarding a competitor's onboarding, a flow that was redesigned four months ago is worse than no reference at all. How do you handle that, do you re-capture flows on a schedule, or does every flow carry a captured-on date so the reader can judge for themselves?