The indexer does not track which ledgers were partially processed due to page-level RPC errors, creating undetectable gaps in the index

[Lost-Z]

Summary

When the indexer's pagination loop encounters an RPC error on a non-first page (e.g. page 3 of a 5-page batch), it exits the loop and advances the cursor to the tip. Events on the failed pages are silently skipped. The error is logged and counted in errors_total, but the specific ledger range that was missed is not recorded. There is no way to identify or backfill the gap without replaying the entire retention window.

Expected Behavior

Partial-page failures should record the affected ledger range so operators can trigger a targeted backfill.

Proposed Solution

• Add a missed_ranges table to the schema: (from_ledger, to_ledger, reason, detected_at)
• When a page fetch fails mid-batch, insert a row into missed_ranges
• Add a GET /health field listing the count of unrecovered missed ranges
• Add a CLI command lumenqraph-indexer recover-gaps that replays all recorded missed ranges

Affected Files

• crates/lumenqraph-indexer/src/poller.rs
• crates/lumenqraph-api/src/routes/health.rs
• migrations/ (new migration)