Reply archive gap recovery marks note replies untruncated and skips tombstones silently #67

Impact: reply-archive gap recovery records syndication data less honestly than the rest of the engine.

1. Records are built with truncated: false even when the payload carries note_tweet. Syndication returns only the preview for long-form posts, so a preview is archived as full text with no truncation flag.
2. A 200 response whose payload is a tombstone (no text, no media) hits if (!text && !(payload && payload.mediaDetails)) return;. It becomes neither an unavailable record nor an entry in errors, so the reply stays "known but uncaptured" with no reason.
3. Text is flattened with .replace(/\s+/g, ' '), so line breaks in replies are lost.

Where: enrichReplyArchiveViaSyndication in sourcecapsule.user.js (~10196-10215). Compare syndicationToQuoteBlock (8216), which sets truncated: !!t.note_tweet, and the GraphQL path, which marks Tombstone/Unavailable typenames unavailable.

Evidence: Node repro with stubbed syndication. A payload with note_tweet produces truncated: false. A TweetTombstone payload produces neither a tombstone record nor an error.

Environment: main 72777ef (v1.6.8). Engine loaded in Node through the module.exports guard. No browser.

Next action: set truncated: !!payload.note_tweet, treat a tombstone typename like the 404 path, and keep line breaks in the stored text.