fix(TransactionClient): don't resubmit repriced (already-mined) replacements by droplet-rl · Pull Request #3504 · across-protocol/relayer

droplet-rl · 2026-06-18T22:11:48Z

Problem

Follow-up to #3503, addressing the Codex P1 raised on that PR (which had already merged ~2 minutes before the review landed).

#3503 opted inventory rebalance transfers into TransactionClient._submit's confirmation loop (ensureConfirmation: true). That loop's TRANSACTION_REPLACED handler resubmits under a fresh nonce on any replacement:

case ethers.errors.TRANSACTION_REPLACED:
  ...
  return this._submit(txn, { nonce: null, maxTries: maxTries - 1 });

ethers (v5) classifies a same-nonce replacement via error.cancelled:

reason: "repriced" → cancelled === false — a gas speed-up of the same logical transaction. The transfer has already mined under a new hash.
reason: "cancelled" / "replaced" → cancelled === true — the original did not take effect.

In the repriced case the current code resubmits a transfer that already landed → double-send (e.g. a duplicated inventory rebalance). This affects every ensureConfirmation caller (rebalancer, refiller, finalizers, dataworker), not just the inventory path.

Fix

In the TRANSACTION_REPLACED branch, return the mined replacement when cancelled === false; only resubmit on genuine cancellations/replacements (cancelled === true). Returning error.replacement also hands the caller the actual mined tx (correct hash) rather than the original, now-replaced one.

A repriced replacement can still mine and revert. ethers attaches the replacement's mined receipt to the error, and waitForTransaction does not throw on revert, so receipt.status === 0 would otherwise be reported as a confirmed success. The cancelled === false branch now checks error.receipt.status and, on a reverted replacement, throws (mirroring the existing CALL_EXCEPTION failure path) so callers don't record a reverted rebalance/finalization as confirmed. (Codex P2.)

Testing

tsc --noEmit, eslint, prettier: clean
test/TransactionClient.ts — 12 passing, including:
- existing Resubmits on TRANSACTION_REPLACED (genuine replacement → still resubmits),
- Does not resubmit a repriced (already-mined) replacement (regression guard: wait() called once, mined replacement returned, no resubmit), and
- new Treats a reverted repriced replacement as a failure (replacement receipt status === 0 → surfaced as a failure, no resubmit).

Docs

No README/AGENTS updates needed — this behavior is documented only in the in-code comment, which is updated here.

🤖 Generated with Claude Code

…cements The ensureConfirmation TRANSACTION_REPLACED handler resubmitted under a fresh nonce on *any* replacement. ethers reports a same-nonce speed-up of the *same* logical transaction as TRANSACTION_REPLACED with cancelled === false; in that case the transaction has already mined under a new hash, so resubmitting double-sends (e.g. duplicating an inventory rebalance transfer). Return the mined replacement instead, and only resubmit on genuine cancellations or replacements (cancelled === true), where the original did not take effect. Surfaced by the Codex review on #3503 (P1) and flagged there as a follow-up. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 48f761feb8

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-06-18T22:14:46Z

+              if (cancelled === false && isDefined(replacement)) {
+                this.logger.warn({
+                  ...common,
+                  message: `Transaction on ${chain} was repriced and already mined; using replacement.`,
+                });
+                return replacement;


Preserve failed replacement receipts as failures

When a repriced replacement is mined but reverts, ethers still reports TRANSACTION_REPLACED with cancelled === false and includes the replacement receipt; this branch returns the replacement as a successful TransactionResponse without checking that receipt status. In that scenario an ensureConfirmation caller records a failed on-chain operation as successful instead of following the existing CALL_EXCEPTION failure path, so reverted rebalances/finalizations can be skipped by the bot.

Useful? React with 👍 / 👎.

Good catch — fixed in 1073374. ethers' waitForTransaction (used internally to fetch the replacement's receipt) doesn't throw on revert, so a repriced replacement that mined with status === 0 was being returned as a confirmed success.

The cancelled === false branch now inspects error.receipt.status: a reverted replacement (status === 0) is thrown, mirroring the CALL_EXCEPTION failure path above, so ensureConfirmation callers no longer record a reverted rebalance/finalization as successful. A successful replacement (status === 1, or an older receipt with no status) is still returned. Added a regression test ("Treats a reverted repriced replacement as a failure").

A repriced replacement (cancelled === false) can mine and revert. ethers attaches the replacement's mined receipt to the TRANSACTION_REPLACED error, and waitForTransaction does not throw on revert, so receipt.status === 0 was previously returned as a confirmed success. Check the receipt status and, when it reverted, throw (mirroring the CALL_EXCEPTION path) so callers don't record a reverted rebalance/finalization as confirmed. Adds a regression test; updates the existing repriced-success test to carry a status === 1 receipt. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

droplet-rl · 2026-06-18T22:21:35Z

Thanks @codex — addressed the P2 in 1073374.

A repriced replacement (cancelled === false) can mine and revert. Since ethers' internal waitForTransaction doesn't throw on revert, the replacement's receipt.status === 0 was being returned as a confirmed success. The cancelled === false branch now inspects error.receipt.status and throws on a reverted replacement (mirroring the CALL_EXCEPTION failure path), so ensureConfirmation callers no longer record a reverted rebalance/finalization as successful. A successful replacement is still returned as before. Added a regression test and updated the PR description.

@codex please re-review the updated diff.

droplet-rl · 2026-06-18T22:22:15Z

Addressed the Codex review on #3504 (commit 1073374).

The single inline P2 ("Preserve failed replacement receipts as failures") was valid: a repriced replacement (cancelled === false) can mine and revert. ethers attaches the replacement's mined receipt to the TRANSACTION_REPLACED error, and its internal waitForTransaction doesn't throw on revert, so receipt.status === 0 was being returned as a confirmed success — letting an ensureConfirmation caller record a reverted rebalance/finalization as successful.

Fix: the cancelled === false branch now inspects error.receipt.status and, on a reverted replacement (status === 0), throws — mirroring the existing CALL_EXCEPTION failure path. A successful replacement (status === 1, or an older receipt with no status) is still returned as before.

Added regression test "Treats a reverted repriced replacement as a failure" (excluded from results, no resubmit); updated the existing repriced-success test to carry a status === 1 receipt. 12 passing.
tsc --noEmit, eslint, prettier: clean.
Replied on the inline thread and posted a top-level comment tagging @codex for re-review; updated the PR description.

droplet-rl · 2026-06-18T22:22:24Z

🔎 View trace

chatgpt-codex-connector · 2026-06-18T22:25:20Z

Codex Review: Didn't find any major issues. More of your lovely PRs please.

Reviewed commit: 107337401c

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

droplet-rl · 2026-06-18T22:25:50Z

Thanks @codex — re-review on 107337401c came back clean (P2 addressed). Nothing further to change here; this is ready for a maintainer's look.

droplet-rl · 2026-06-18T22:25:55Z

🔎 View trace

chatgpt-codex-connector · 2026-06-18T22:26:41Z

Summary

No code changes were made. The trigger says the re-review on 107337401c came back clean and that nothing further is needed, so I did not create a new commit or PR.
No README/AGENTS updates are needed because there is no behavior, configuration, interface, or runtime-flow change in this follow-up. AGENTS.mdL13-L20

Testing / Checks

✅ pwd && find .. -name AGENTS.md -print && git status --short
✅ cat AGENTS.md && git diff -- yarn.lock | sed -n '1,120p'
✅ nl -ba AGENTS.md | sed -n '1,24p'

Note: git status --short shows an existing modified yarn.lock; I did not alter or commit it.

View task →

droplet-rl · 2026-06-18T22:27:48Z

Confirmed — no further action. The PR diff is limited to src/clients/TransactionClient.ts and test/TransactionClient.ts; the modified yarn.lock is a local working-tree artifact and is not part of this branch/commit. PR stands as-is, ready for a maintainer's review.

droplet-rl · 2026-06-18T22:27:53Z

🔎 View trace

droplet-rl requested review from bmzig, dijanin-brat, mrice32, nicholaspai and pxrl as code owners June 18, 2026 22:11

droplet-rl mentioned this pull request Jun 18, 2026

fix(rebalancer): wait for inventory rebalance tx confirmation #3503

Merged

chatgpt-codex-connector Bot reviewed Jun 18, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fix(TransactionClient): don't resubmit repriced (already-mined) replacements#3504

fix(TransactionClient): don't resubmit repriced (already-mined) replacements#3504
droplet-rl wants to merge 2 commits into
masterfrom
droplet/fix-transaction-replaced-repriced

droplet-rl commented Jun 18, 2026 •

edited

Loading

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Jun 18, 2026

Uh oh!

droplet-rl Jun 18, 2026

Uh oh!

droplet-rl commented Jun 18, 2026

Uh oh!

droplet-rl commented Jun 18, 2026

Uh oh!

droplet-rl commented Jun 18, 2026

Uh oh!

chatgpt-codex-connector Bot commented Jun 18, 2026

Uh oh!

droplet-rl commented Jun 18, 2026

Uh oh!

droplet-rl commented Jun 18, 2026

Uh oh!

chatgpt-codex-connector Bot commented Jun 18, 2026

Uh oh!

droplet-rl commented Jun 18, 2026

Uh oh!

droplet-rl commented Jun 18, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Uh oh!

Conversation

droplet-rl commented Jun 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Problem

Fix

Testing

Docs

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Jun 18, 2026

Choose a reason for hiding this comment

Uh oh!

droplet-rl Jun 18, 2026

Choose a reason for hiding this comment

Uh oh!

droplet-rl commented Jun 18, 2026

Uh oh!

droplet-rl commented Jun 18, 2026

Uh oh!

droplet-rl commented Jun 18, 2026

Uh oh!

chatgpt-codex-connector Bot commented Jun 18, 2026

Uh oh!

droplet-rl commented Jun 18, 2026

Uh oh!

droplet-rl commented Jun 18, 2026

Uh oh!

chatgpt-codex-connector Bot commented Jun 18, 2026

Summary

Uh oh!

droplet-rl commented Jun 18, 2026

Uh oh!

droplet-rl commented Jun 18, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

droplet-rl commented Jun 18, 2026 •

edited

Loading