[clement-ux]

Why

ousdUsdeV4Rebalance (every 2 min, --network mainnet --cap-bps 1) failed 14 of 687 runs in 24h (~2%), all in currentTokenId(): module.queryFilter(Rebalanced(poolId)) over 50,000-block chunks, no retry. The RPC is the dRPC load balancer, which routes each call to one of several upstreams:

• 12× Temporary internal error. Please retry (HTTP 500, JSON-RPC code 19)
• 1× query block range exceeds server limit, narrow your filter: 10000 (-32602)
• 1× block range extends beyond current head block (lagging upstream)

A failed run sends nothing and the next one succeeds, but each failure fires a Talos alert.

What changed

New tasks/lib/logs.ts → findLatestLog() (no retry / chunked-getLogs helper existed in tasks/lib/), used by both ousdUsdeV4Rebalance and ousdSusdeV4Rebalance (same lookup):

• Chunks ≤ 9,999 blocks (to - from + 1 <= 10000). --log-chunk default is now 9999 and any larger value is capped.
• Head queried on its own: the newest 5 blocks are a separate first query, so only that tiny range can hit a lagging upstream. It is retried, not skipped: a Rebalanced there is our own run from 2 min ago, and missing it would read the burned position (the run would then fail on ownerOf).
• Retry: 3 attempts per range, backoff 500 ms → 1 s, on transient errors (5xx / SERVER_ERROR, "Temporary internal error", "beyond current head", timeouts). Non-transient errors (e.g. revert) throw immediately.
• Range-limit error halves the chunk and retries the same to instead of failing.
• Retries log at warn; docs/ACTIONS.md documents --lookback-blocks / --log-chunk for both actions.

Tests

tasks/test/logs.js (8 cases): the three dRPC errors are classified; every query ≤ 10,000 blocks and contiguous down to the floor; event in the head range is found; retry then success (backoff 100/200); range-limit halves 9999 → 4999 → 2499 → 1249; throws after retries are exhausted; no retry on a non-transient error.

• pnpm test:tasks: 157 passing
• pnpm prettier:ts, eslint on touched files: clean

Mainnet dry run

$ DEPLOYER_PK=<throwaway> pnpm exec tsx tasks/run.ts ousdUsdeV4Rebalance --network mainnet --cap-bps 1 --dryrun
[ousdUsdeV4Rebalance] Running on mainnet (1)
[ousdUsdeV4Rebalance] Latest Rebalanced at block 26235102: minted 474867
[ousdUsdeV4Rebalance] Ethena redeems 50000 USDe at 0.999979 USDC (0.21 bp)
[ousdUsdeV4Rebalance] KyberSwap: 0.999873 USDC/USDe, exit 0.999881 USDC/OUSD; target 0.999879 OUSD/USDe (-1.21 bp vs par), capped at the Ethena redemption value - 1 bp -> tick 2, top bid 0.999800
[ousdUsdeV4Rebalance] Position 474867 [2, 3]: skipping (In place, holds 29860.765771475934593322 OUSD, 7.891655951176555697 USDe)
[ousdUsdeV4Rebalance] Completed in 2.6s

ousdSusdeV4Rebalance --dryrun also finds its event ~39.5k blocks back (several 9,999-block chunks): Latest Rebalanced at block 26200340: minted 463219. (Its simulation then reports Contract with a Signer cannot override from with a non-operator throwaway signer — untouched by this PR.)

🤖 Generated with Claude Code