This is a finishing pass, not another feature pass.

The product is not complete until the following are true:

1. Fix the self-transfer bug:

   • from == to must not close a holding streak
   • must not set dropped_to_zero
   • must not mark the wallet as having transferred out
   • add and pass the existing regression test

2. Fix historical seat-source resolution:

   • resolve the ChadBoard seat NFT contract at the declared snapshot as-of block
   • do not use the current contract configuration for older snapshots
   • ensure the seat event index is synced through the snapshot block
   • add and pass the existing bounded-backfill regression test

3. Run the complete historical backfill.

   • Do not claim completion based on samples.
   • If the current RPC’s 10-block limit makes this impractical, measure the actual rate and report it.
   • Use adaptive batching, retries, and checkpoints.
   • Do not silently omit blocks or use incomplete rankings.

4. Deploy the backend service on the IMD worker/VPS.

   • Expose the API over HTTPS.
   • Keep the RPC key server-side.
   • Never include it in this prompt, source, logs, frontend assets, or documentation.

5. Connect the deployed frontend to the hosted API.

   • No manual #api URL should be required for normal visitors.
   • Keep Radar enrichment optional but make the Radar panel functional when the service is available.

6. Run a real end-to-end production check:

   • /v1/status
   • /v1/site-feed
   • wallet lookup
   • leaderboard
   • retention ranking
   • target filters
   • frontend loading the live snapshot

7. Do not mark the work complete until the public site no longer says: “Awaiting the first snapshot”

   and displays real snapshot metadata:

   snapshot ID, as-of block, observed timestamp, total seats, activated seats, and ranked wallet count.

8. If the RPC cannot complete the historical backfill, stop and report the exact blocker instead of publishing partial rankings.