fix(filler): defer the eager missing-mint reconciliation while the reader is catching up

[igorls]

Problem

AtomicAssetsHandler.init() loops CALL update_atomicassets_mints (bounded 50k/call) + a full atomicassets_assets COUNT until the missing-mint backlog is under 50k. While the reader is catching up the update_atomicassets_mints job is deferred (shouldDeferDrain), so the backlog grows large — ~8M rows observed on WAX mainnet.

Running this reconciliation at boot then:

• blocks the reader from ever starting (it loops ~160× on an 8M backlog), and
• busts the connection statement_timeout on the 466M-row COUNT,

so any filler restart mid-catchup becomes an init crash-loop — the init promise rejects and the reader never starts (the process stays up but idle).

Fix

Gate the reconciliation on the reader's distance from head. shouldDeferDrain() can't be used here because reader.blocksUntilHead is still 0 before the reader starts, so compare the reader's persisted contract_readers.block_num to chain head directly and skip while far behind (>10k blocks). The deferred update_atomicassets_mints job fills the backlog once the reader reaches head. The reconciliation loop body is unchanged.

Validation

• pnpm check-types clean.
• On the WAX-mainnet cluster: a filler restart mid-catchup now logs Reader is ~N blocks behind head - skipping eager missing-mint reconciliation and the reader resumes immediately instead of crash-looping.

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

Defers the expensive “missing mint number reconciliation” work during startup when the reader is far behind chain head, avoiding long startup stalls and potential crash-loops from DB statement_timeout.

Changes:

• Adds a reader-lag gate by comparing the persisted reader position (contract_readers) to chainInfo.head_block_num.
• Skips eager mint reconciliation when the reader is >10k blocks behind head, logging the deferral.
• Avoids querying atomicassets_config entirely when reconciliation is skipped.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[robrigo]

Review (FACINGS) — LGTM.

Verified chainInfo (get_info()), this.filler.reader.name, and positiveIntEnv all exist. The gate fails closed — unknown/non-numeric position → null → skip — which is the safe direction (never falls open into the expensive reconciliation during catch-up). Patch applies cleanly onto feat/atomicassets-v2.

Note: Copilot's three inline comments were on an earlier commit — all three are already addressed in the current head: the threshold is the env-tunable MINT_RECONCILIATION_MAX_LAG_BLOCKS constant, Number.isFinite + clamp handle the NaN case, and the skip branch uses a plain [] rows array (not {rows: [] as any[]}).

[igorls]

Rechecked against current main: the startup missing-mint reconciliation in src/filler/handlers/atomicassets/index.ts still runs ungated — no reader-lag check around the get_info() / atomicassets_config count-and-update_atomicassets_mints loop — so the catch-up startup stall this defers is still present. Branch is MERGEABLE and the gate fails closed (unknown/non-numeric reader position → skip), so it never falls open into the expensive reconciliation during catch-up. Ready to merge.