Fix: In-flight Transaction Fails After a Governance Update

Symptom

A transaction that was pre-flight checked successfully is later rejected at the miner level. The error from ckb-cli or your broadcast client is a “Dead reference” or “TransactionFailedToResolve” error indicating a consumed cell dep.

Cause

Between when you built the transaction (with a registry cell dep) and when it was mined, a governance update confirmed. The governance transaction consumed the registry cell your transaction referenced as a dep. From the miners’ perspective, your transaction’s cell dep no longer exists.

Fix

Rebuild the transaction with the current registry cell outpoint:

1. Discover the new outpoint:

   const { txHash, index } = await findRegistryCell(rpcUrl, registrySpec);

2. Re-fetch the registry payload:

   const registry = await fetchRegistryPayload(rpcUrl, txHash, index);

3. Re-run the pre-flight check with the updated registry:

   const result = preflightCheck(outputs, [registry]);

4. Re-build the transaction with the new cell dep outpoint and re-sign.

If the address you were sending to was not blacklisted before but is now — because the governance update added it — the pre-flight check will catch that before you re-sign.

Prevent future occurrences

This failure mode is inherent to the single-cell registry model — one governance update invalidates all in-flight transactions that reference the old cell. Mitigations:

• Subscribe to registry updates. If your application architecture supports it, listen for governance transactions and proactively rebuild any pending transactions when the registry cell changes.
• Keep the time between build and broadcast short. The longer a transaction sits unsigned, the higher the chance a governance update occurs in that window.
• Coordinate governance updates. Registry operators should announce upcoming governance updates and submit them at low-traffic periods to minimize the number of affected in-flight transactions.

See Registry design for a full discussion of the single-cell model and its implications.