LND adds ceiling for external Bitcoin fee estimates at 10,000 sat/vB

3 min

LND’s October 22 release candidate v0.21.5-beta.rc1 adds guardrails around external fee estimates to reduce wallet risk and repairs a native SQL connection-pool deadlock. The build is pre-release and requires no database migration; signature and binary verification are provided, while real-world testing remains necessary.

TL;DR

Published build: v0.21.5-beta.rc1 pre-release.
New fee ceiling: 10,000 satoshis per vbyte on external API inputs, plus bounds on minimum-relay estimates.
Verification confirms download integrity; deployment testing confirms behavior.

Remote estimates need limits before they reach the wallet

The release notes document that external fee inputs previously had only a lower bound. The repair applies a ceiling before estimates can affect fee-sensitive wallet operations.

The change sets an upper limit of 10,000 satoshis per virtual byte and bounds the minimum-relay value from the API. The cap functions as an extreme-value guardrail, not guidance on typical fees or current network requirements.

Fee estimates are inputs to decisions. External services can return unusable values due to errors, data format changes, or compromised dependencies. Bounding inputs limits how upstream problems can influence a wallet.

This aligns with the need for processes that find and fix issues in money-handling software, as outlined in Bitcoin Core’s revised disclosure process.

A candidate build needs evidence from its deployment setting

The GitHub page marks the build as pre-release and notes no database migration. Signature and binary checks confirm what was downloaded, not behavior across all node configurations.

The notes include a fix for a native SQL connection-pool deadlock. Node reliability depends on storage as well as network communication. A process can stay running while requests stop completing, which still fails dependent applications.

Testing should reflect actual service work: receiving requests, accessing storage, handling failures, and restarting cleanly. Removing a database migration eliminates one change category, not all interactions with existing data.

Payment operations differ from custody. Shared-key Bitcoin lending, such as shared-collateral keys, focuses on signing authority, while payment nodes focus on availability and correctness. Custody and recovery arrangements do not replace software maintenance. This candidate introduces specific changes to evaluate; the final release and each operator’s deployment remain separate decisions.

Written by the News Desk and edited by Samuel Rae.