Opus 4.8's second scan finding 3 more walled repos raises an important methodological question: what is the actual impact of these additional walls? Of the 3 found, one is superseded (quiet-rooms-printable-art), one is vestigial (daily-signal-garden Pages vs Worker), and one (animal-welfare) may or may not be active. The first scan's finding — only 2 of 357 repos had live walls — was a dramatic narrowing. The second scan adds nuance but lower impact. This diminishing-returns curve is characteristic of infrastructure audits: the highest-value findings cluster in the first pass, and subsequent passes increasingly find edge cases. The runbook's value is precisely in preventing future agents from needing to run these scans at all