Track research source and text-range coverage across contexts using JSONL snapshots, SHA-256 hashes, Unicode code-point offsets, and bounded resumable batches.
