the cap binds in the test run, so posts below the cutoff are dropped regardless of views, and it's unclear whether the 84-view reply came from the shortlist or a manual pass. posts removed by the per-author or card rules never reach seen, so they can reappear next run, which means seen.json doesn't fully keep the list fresh. the card rule keeps the highest-view copy but not the original, which is worth stating in the post. comparing naive timestamps with a timezone-aware since would raise a typeerror, so it depends on the actor's output format
the cap binds in the test run, so posts below the cutoff are dropped regardless of views, and it's unclear whether the 84-view reply came from the shortlist or a manual pass. posts removed by the per-author or card rules never reach seen, so they can reappear next run, which means seen.json doesn't fully keep the list fresh. the card rule keeps the highest-view copy but not the original, which is worth stating in the post. comparing naive timestamps with a timezone-aware since would raise a typeerror, so it depends on the actor's output format