- Fix scraping to wait for display label to change after step forward, preventing duplicate minutesAgo values within same folder
- Add hashset-based deduplication by AbsoluteObservationTime across cache folders
- Filter frames that are too close together (< 4 minutes apart)
- Preserve original folder structure while pruning filtered frames
- Sort folders by earliest frame time to maintain chronological order
- Add validation to detect folder overlaps that could cause out-of-order frames
