This week’s newsletter summarizes an analysis of decentralized mining pool payouts, describes the disclosure of a fixed denial-of-service vulnerability in Eclair, and links to continued discussion about whether stale tip relay should include block data. Also included are our regular sections describing recent changes to services and client software, announcing new releases and release candidates, and describing notable changes to popular Bitcoin infrastructure software.

News

  • ● Framework for decentralized mining pool payouts: Developer 7d5x9 posted to Delving Bitcoin an analysis of how decentralized mining pools can pay miners directly from coinbase transactions, which the post calls decentralized coinbase payout protocols (DCPPs). Giving every miner its own output does not scale because mining firmware limits coinbase transaction size and each output displaces fee-paying transactions. The post argues that when a pool has more miners than outputs, any rule ranking miners by work allows a rich list attack, where a large miner splits its hashrate across identities that each still outrank the smallest miners, pushing them out of the coinbase transaction.

    According to 7d5x9, earlier designs only move this problem. Braidpool shifts it from who receives an output to who controls the pool’s threshold key (see Newsletter #340). For a single coinbase output paying a joinpool with unilateral exits, the obstacle is that miners vary the coinbase transaction for extra nonce space, so its txid is unknown and exit transactions cannot be signed until a block is found. 7d5x9 believes this leaves the output controlled by a single key or threshold of keys until a soft fork enables rebindable signatures, such as those proposed in BIP448 (see Newsletter #397). In sketches based on Spark-style statechains and on Ark, an attacker keeping one small claim per subtree can block cheap cooperative settlement, raising miners’ costs, though the post considers this less severe for Ark.

    The post concludes that a joinpool does not decide which outputs enter the coinbase transaction. It sees two options for that: a value-weighted random draw (an idea credited to Matt Corallo) or outputs reserved for designated operators.

    LearnBitcoin.com, who ran a public P2Pool node, examined 2,545 P2Pool coinbase transactions and argued that growth alone, without any attacker, squeezed out small miners because P2Pool raised the share difficulty of miners expected to earn under 0.001 BTC per block. For its successor P2Poolv2, jungly suggested a limit of 16 outputs given current firmware. Steven Roose replied that he could not think of a scheme worth exploring without covenants, which would let a coinbase output deterministically commit to an Ark transaction tree that pays every miner who uses the same Ark server.

  • ● Disclosure of a DoS vulnerability in Eclair: rorp posted to Delving Bitcoin the responsible disclosure of a memory-exhaustion denial-of-service (DoS) vulnerability affecting Eclair version 0.14.0 and fixed in July 2026 in release 0.14.1. Any peer able to connect to a vulnerable node could crash it, even without having a channel with it. The bug gave an attacker no direct access to funds, but repeatedly crashing a node could put its channel funds at risk. Users still running 0.14.0 should upgrade.

    The attack vector leveraged a missing size check in Eclair. According to BOLT2, messages can be sent in batches to be treated as a single logical update through the start_batch message, which among other fields carries the batch_size. According to the specifications, the size must be at most 20 and strictly greater than 1. While Eclair checked for the upper bound, it did not enforce the lower one.

    An attacker would have been able to use this vulnerability by sending a start_batch message with batch_size = 0, followed by commitment_signed messages, which are sent in batches during splicing. Since the code stopped adding messages to its internal batch only once the number of received messages reached the batch_size (which was 0, thus always lower than the number of messages received), the list never stopped growing, leading to memory exhaustion. The vulnerability was fixed by having the code check the lower bound, too, as specified in BOLT2.

  • ● Continued discussion of stale tip relay: Antoine Poinsot replied to the Bitcoin-Dev mailing list thread about BIP332 stale tip relay (see Newsletters #417 and #423) to suggest that nodes initially relay only the headers of stale blocks. Poinsot argued that headers are enough to make stale blocks observable. Relaying the block data as well lets nodes reorganize faster to a chain built on a stale block, which increases the expected revenue of a miner that briefly keeps mining on its own block after losing a block race. Poinsot considered that effect most likely small but not worth working out rigorously.

    Anthony Towns replied that any miner who would benefit can cheaply implement stale block relay itself. Towns estimated that a pool with 20% of network hashrate already gains about 16.1 BTC per year by mining on its own block instead of a competing block it saw first at the same height, and that cutting the relay delay of the resulting one-block reorg from five seconds to one would raise that to 16.36 BTC. Towns also argued that stale headers can raise concerns, such as possible selfish mining, but only the block data can settle them. Poinsot agreed that the estimate appears to cover the effect on honest mining incentives and, while still seeing most of the value in headers, did not want to block relaying full blocks.

Changes to services and client software

In this monthly feature, we highlight interesting updates to Bitcoin wallets and services.

  • ● Liana v16.0 released: Liana v16.0 writes its encrypted descriptor backups in the BIP138 format (see Newsletter #424) and makes taproot the default for new wallets.

  • ● COLDCARD firmware adds codex32 backups: Coinkite announced COLDCARD firmware versions 5.6.3 and 1.5.3Q, which add support for generating and importing codex32 secrets and for splitting the device’s wallet into two to nine Shamir shares for later recovery. The releases also add support for per-input height and time locktimes in PSBTv2 transactions.

  • ● Wavelength released on mainnet: Lightning Labs announced that Wavelength, its Ark-like payments toolkit (see Newsletter #415), is available on mainnet as an alpha with accounts limited to 300,000 sats.

  • ● Bull Bitcoin makes payjoin the default for exchange orders: Bull Bitcoin announced that onchain purchases and sales made on its exchange through the Bull Bitcoin Mobile wallet now use payjoin by default (see Newsletter #338). Version 6.14.0 of the wallet also adds experimental support for sending and receiving silent payments.

  • ● Breez SDK adds silent payment sending: Version 0.28.0 of the Spark-based Breez SDK adds support for sending to silent payment addresses.

  • ● Silent payments scanner for Electrum plugin: Ali Sherief announced spscand, a watch-only silent payments scanner that works with the Electrum sending plugin covered in Newsletter #419. It can scan for payments using the user’s own Bitcoin Core node, a BlindBit server, or a BlindBit server combined with compact block filters from P2P peers.

  • ● Stratum v2 releases and pool adoption: The Stratum v2 reference implementation released sv2-apps v0.9.0, which adds support for Bitcoin Core 32.x, including submitting found blocks over its IPC interface, and hardens its Job Declarator Client (see Newsletter #384 for previous coverage). Separately, MARA Foundation announced that MARA mined its first block using Stratum v2, and Foundry announced that it has begun testing Stratum v2 on its pool.

  • ● Coinswap renamed to OpenSwap: OpenSwap v0.2.3 is the first release under the new name of the coinswap protocol implementation formerly called Coinswap (see Newsletter #415). It adds an Electrum server backend option, swaps that pay a third-party address, and, when built with Lightning support, swaps between onchain funds and LN through LDK Server (see Newsletter #406).

Releases and release candidates

New releases and release candidates for popular Bitcoin infrastructure projects. Please consider upgrading to new releases or helping to test release candidates.

  • ● Core Lightning 26.12rc1 is a release candidate for the next major version of this popular LN node implementation. It will be the first major release since 26.06 and covers two release cycles. It adds a tool for recovering HTLC outputs after channel-state loss (see Core Lightning #9730 below), improves payment pathfinding (see Newsletters #418 and #430), and adds experimental support for a new cooperative close protocol (see Newsletter #413). The release notes also announce that xpay will replace pay in 27.03.

Notable code and documentation changes

Notable recent changes in Bitcoin Core, Core Lightning, Eclair, LDK, LND, libsecp256k1, Hardware Wallet Interface (HWI), Rust Bitcoin, BTCPay Server, BDK, Bitcoin Improvement Proposals (BIPs), Lightning BOLTs, Lightning BLIPs, Bitcoin Inquisition, and BINANAs.

  • ● Bitcoin Core #36757 fixes a privacy leak in address discovery over Tor and I2P. Previously, when a node connected outbound to a malicious Tor or I2P peer, that peer could claim to see the node at an arbitrary routable IP address in its version message. Bitcoin Core could then advertise that address back to the peer, along with its own clearnet listening port. A distinctive port could then help correlate the connection with a clearnet node that was observed using the same port (see Newsletter #52 for an earlier Tor listening-port fix). Bitcoin Core now ignores the address that Tor and I2P peers report seeing the node at.

  • ● BIPs #2356 replaces the placeholder genesis parameters in BIP95 with those of the mined testnet5 genesis block and marks the BIP as Deployed. The genesis coinbase includes the date 2026-11-14 and the hash of mainnet block 975,744. Committing to that hash ensures that the testnet5 genesis block could only be mined after mainnet reached that height (see Newsletter #431).

  • ● Core Lightning #9730 adds the lightning-hsmtool recoverhtlc command to manually recover HTLC outputs after local channel-state loss. Once the peer publishes its latest commitment transaction, the recovering node can already reclaim its main output using a static channel backup (see Newsletter #421). Given the local hsm_secret, channel and HTLC details, and the peer’s per-commitment point, the new command identifies matching HTLC outputs. The tool constructs signed transactions to reclaim outgoing HTLCs after a timeout or to claim incoming HTLCs with known preimages. However, the operator must manually review and broadcast the transactions. A cooperative CLN peer can export the public point using the dumpcommitments --point-only command without revealing secrets.

  • ● LDK #5561 fixes a potential loss of forwarded funds when a peer sends revoke_and_ack before LDK sends the corresponding commitment_signed message. Previously, LDK could accept this premature acknowledgment and advance its internal commitment number. Consequently, it could sign a commitment that differed from the one recorded by its onchain monitor. If the peer published that commitment, LDK might fail to claim the incoming HTLC, even after paying the outgoing HTLC, despite knowing its preimage. LDK now rejects the premature acknowledgment and force closes the channel. The fix was released in LDK 0.2.8 (see Newsletter #431). Newsletter #426 described an earlier bug that left a signed commitment unmonitored during reconnection.

  • ● LDK #5562 and #5545 fix channel ID collisions that a malicious peer could create between its channels and an LDK node. In the first, the peer could reuse a force-closed channel’s known ID as a new channel’s temporary ID. This could prevent LDK from recording a forwarded payment’s preimage in the closed channel’s monitor, leaving the incoming HTLC unclaimed after the outgoing HTLC had been paid. In the second, the peer could calculate a channel’s future final ID from its funding transaction and submit another opening request with that temporary ID. If the original channel adopted its final ID before the application accepted the new request, acceptance could overwrite its funded state. LDK now rejects conflicting IDs and rechecks them at acceptance. Both fixes were also released in LDK 0.2.8. See Newsletter #427 for related channel ID collisions that could cause a DoS in Eclair.

  • ● LDK #5557 fixes a vulnerability that allowed malicious peers to disrupt transaction monitoring on nodes using an Electrum backend. The peer could propose a channel whose claimed funding output reused an already-watched outpoint, but with a different locking script. Registering that proposal could overwrite the legitimate watch’s stored script, even if LDK subsequently rejected the channel. LDK would then query Electrum for transactions associated with the bogus script and might miss a transaction spending the real output, preventing a timely response to a channel closure. LDK now retains every distinct script registered for an outpoint, so a bogus registration cannot suppress the legitimate watch. This fix was also released in LDK 0.2.8.

  • ● LDK #5558 and #5482 fix two ways payments could exceed their fee budget. The first PR addresses the handling of an omitted fee-limit setting for async payments using static BOLT12 invoices (see Newsletter #321). LDK now applies its default cap of 1% of the payment amount plus 50 sat when the application omits the setting. Previously, omitting it removed the limit entirely, allowing a malicious receiver to supply a blinded path with excessive fees. This fix was also released in LDK 0.2.8. The second PR addresses retries when a route adds value to meet an HTLC minimum. LDK treated the extra value as part of the requested payment and restored the failed path’s fee budget on retry, allowing the total spend to exceed its limit. LDK now retries only the remaining amount owed and counts any excess against the remaining fee budget.

  • ● LND #11555 changes how LND reports non-AMP invoices as settled. Although applications could already monitor when the settlement of individual HTLCs was locked in offchain (see Newsletter #224), LND could mark an invoice as settled as soon as the preimage was released. The invoice now enters the PENDING_SETTLE state on preimage release. Once every HTLC in the payment has a final offchain or onchain outcome, LND marks the invoice as SETTLED if all parts were collected or CANCELED if any failed. A canceled invoice can still have collected some parts, which are not refunded, so applications should credit only the amounts from finalized, successful HTLCs. An invoice’s preimage_released flag records whether LND has committed to releasing the preimage. AMP payment sets still settle at preimage release.

  • ● LND #11551 fixes expiry validation when completing multipath payments and AMP payments. Previously, LND checked each HTLC when it arrived. If more blocks were mined while the receiver waited for the remaining parts, LND could release preimages even though an earlier HTLC no longer had the required expiry margin for onchain collection. Now, LND rechecks every part against the current block height when the final part arrives. If a previously accepted part is now too close to expiry, LND fails that set and leaves the invoice open for another attempt (see Newsletter #424 for earlier work on canceling individual AMP sets).

  • ● LND #11472 and #10842 fix cases where temporary backend errors or one problematic input could block otherwise valid sweeps. The first keeps inputs available for retry after temporary or unclassified errors during mempool acceptance checks (see Newsletter #373). It was released in LND 0.21.8-beta and 0.20.8-beta (see Newsletter #431). The second uses diagnostic transactions that are never broadcast to test subsets of a rejected batch. If an invalid or conflicting input can be identified, LND excludes it from the next attempt and retries the remaining inputs.

  • ● LND #11542 fixes recovery of forwarded HTLCs after channel reconnections. When replaying stored forwarding batches, LND could filter out acknowledged HTLCs and then use the remaining entries’ new positions to look up metadata indexed by their original positions. This could cause unresolved HTLCs to be skipped or misassociated. Additionally, if a channel’s HTLC handler stopped after LND recorded the batch but before handing off an HTLC, an incomplete record could be left behind, causing retries to be discarded as duplicates. LND now preserves the original batch indexes and removes incomplete records after a failed handoff, allowing unresolved HTLCs to be retried. The fix was released in LND 0.21.8-beta and 0.20.8-beta (see Newsletter #431).

Want more?

For more discussion about the topics mentioned in this newsletter, join us for the weekly Bitcoin Optech Recap on Riverside.fm at 17:30 UTC on November 24. The discussion is also recorded and will be available from our podcasts page.