docs: add CCTP receiveMessage fee guidance

[westkite1201]

Summary

Adds a CCTP v2 troubleshooting note for MessageTransmitterV2.receiveMessage() relays from Arc Testnet to destination chains.

The doc explains why viem/MetaMask submissions can fail with:

RPC submit: max fee per gas less than block base fee

and shows how to estimate fees on the destination chain immediately before receiveMessage(), then add maxFeePerGas headroom before calling walletClient.writeContract().

Fixes #153.

Changes

• Adds docs/cctp-v2-receive-message.md
• Documents the stale EIP-1559 fee-estimate failure mode
• Adds a viem helper example for destination-chain fee estimation and maxFeePerGas padding
• Includes an integration checklist for relayers

Duplicate check

• Duplicate gate: PASS
• Artifact: .context/pre-pr-gates/circlefin-arc-node-153-20260621T142653Z.md
• Manual search found no existing PR for bug: CCTP v2 receiveMessage() on destination chains fails with 'max fee per gas less than block base fee' when using default viem gas estimation #153 or the root-cause terms:
  • max fee per gas less than block base fee
  • receiveMessage estimateFeesPerGas
  • CCTP v2 receiveMessage
• Existing PR docs: add MetaMask integration guide with USDC token setup #100 only overlaps README/docs generally; this PR does not touch README and covers a different issue.

Testing

npx prettier --check docs/cctp-v2-receive-message.md
All matched files use Prettier code style.

git diff --check origin/main...HEAD
passed

[osr21]

Thanks for documenting this — the fee-race failure mode on receiveMessage() is a genuine pain point and a 130%-bumped estimate-then-submit pattern is the right call.

A few technical points worth addressing before merge:

---

1. maxFeePerGas ?? fees.gasPrice assigns the wrong type on legacy chains

const maxFeePerGas = fees.maxFeePerGas ?? fees.gasPrice;
// then always returns: { maxFeePerGas: bumpFee(maxFeePerGas) }

On a legacy-gas chain, estimateFeesPerGas() returns { gasPrice: ... } with maxFeePerGas == null. The fallback stores the gasPrice value in a variable named maxFeePerGas, then spreads it as the EIP-1559 field. Many nodes reject transactions that carry maxFeePerGas on a chain that hasn't enabled EIP-1559 — so the bump actually makes things worse on those chains. Consider branching:

if (fees.maxFeePerGas != null) {
  return {
    maxFeePerGas: bumpFee(fees.maxFeePerGas),
    ...(fees.maxPriorityFeePerGas != null
      ? { maxPriorityFeePerGas: fees.maxPriorityFeePerGas }
      : {}),
  };
} else if (fees.gasPrice != null) {
  return { gasPrice: bumpFee(fees.gasPrice) };  // legacy chain
} else {
  return {};
}

---

2. Arc Testnet as destination is a gap in the guide

The doc covers the case where Arc is the source (burns land on Sepolia/Base/Fuji). But the reverse direction — Sepolia → Arc Testnet — hits a different failure mode: Arc uses a legacy gas model, eth_gasPrice is documented to return stale values (see arc-node#87), and Arc nodes reject transactions that carry maxFeePerGas/maxPriorityFeePerGas fields. The mitigation is the gasPrice branch above plus the same 130% bump. Worth adding a brief note specifically for Arc-as-destination since it's the most common cross-direction for this testnet.

---

3. maxPriorityFeePerGas is not bumped

The tip is forwarded as-is from the estimate. On low-tip chains this is harmless, but if the priority fee market spikes concurrently with the base fee, a stale tip can also cause the tx to sit unconfirmed. A modest bump (e.g. 110-120%) on the priority fee as well would be more robust and is consistent with the "add headroom" philosophy of the rest of the guide.

---

Minor note: the ceiling-division in bumpFee ((value * bps + BASIS_POINTS - 1n) / BASIS_POINTS) is correct and the right choice for a fee floor — nice.

[osr21]

The update addresses all three points from the previous review — the EIP-1559/legacy branching is now correct, the Arc-as-destination section is a solid addition, and the priority-fee bump is a good defensive change. A few remaining observations on the updated code:

---

1. return {} is a silent no-op — the protection disappears without the caller knowing

When neither fees.maxFeePerGas nor fees.gasPrice is set the helper returns {}. Spreading that into writeContract leaves it with no fee override, so viem runs its own internal fee estimation — the exact stale-fee path the helper was supposed to prevent. A caller has no way to detect this; the relay appears to use bumped fees but actually uses none.

Consider throwing instead of returning silently:

// instead of: return {};
throw new Error(
  `estimateFeesPerGas returned no usable fee fields on ${destinationChain.name} — cannot construct safe fee overrides`
);

This forces the caller to handle the exceptional case explicitly (e.g., retry after a short delay) rather than silently regressing to the unsafe default path.

---

2. Arc gasPrice staleness is not limited to rapid transaction sequences

The new section reads:

Arc Testnet gas estimates can also be stale during rapid transaction sequences

Issue #87 affects individual transactions too — the Arc RPC's gasPrice response can lag even when no prior transactions have been submitted recently. receiveMessage is typically a single standalone call, so the "rapid sequence" framing undersells the risk for this specific use case. Suggest:

Arc Testnet gas estimates can be stale even for a single transaction; re-estimate gasPrice immediately before submission and apply the same 130% headroom.

---

3. estimateFeesPerGas() is not wrapped — throws propagate to the caller

If the destination RPC is temporarily unreachable, estimateFeesPerGas() throws and the relay fails hard. For a checklist note or a small addition to the helper:

// at the top of estimateReceiveMessageFees:
let fees: Awaited<ReturnType<typeof destinationClient.estimateFeesPerGas>>;
try {
  fees = await destinationClient.estimateFeesPerGas();
} catch (err) {
  throw new Error(`fee estimation failed on ${destinationChain.name}: ${err}`);
}

Wrapping it with a contextual message makes the failure origin obvious in relay logs, especially when a relayer monitors multiple destination chains.

---

On the math: bumping maxFeePerGas by 130% and maxPriorityFeePerGas by 115% always preserves the EIP-1559 invariant (maxFeePerGas ≥ baseFeePerGas + maxPriorityFeePerGas) — verified. The ceiling division in bumpFee is still correct.

[osr21]

All three points from the previous follow-up are addressed:

• return {} replaced with a named throw — callers now get an explicit signal when fee estimation produces no usable fields, rather than silently receiving no overrides
• estimateFeesPerGas() wrapped in try/catch with the chain name in the message — failure origin is now unambiguous in relay logs across multiple chains
• Arc staleness wording corrected — "stale even for a single transaction" is accurate and no longer undersells the risk

Doc looks solid. No further issues from this review thread.