Why This Matters
For most of Bitcoin's history, the mempool — the waiting room where unconfirmed transactions sit before being mined — has operated under a set of ancestor and descendant limits that date back to 2015. Those limits were designed for a simpler era. As transaction volumes grew and fee markets became more sophisticated, the old architecture created blind spots: miners couldn't always build the most profitable blocks, replace-by-fee (RBF) logic could be gamed, and child-pays-for-parent (CPFP) had awkward exceptions bolted on after the fact.
Bitcoin Core 31.0, now in its fourth release candidate as of April 11, changes the foundation. The mempool has been rebuilt from scratch around a concept called cluster mempool — and it brings a separate privacy upgrade that gives node operators a new way to broadcast transactions without exposing their IP address.
What Cluster Mempool Actually Does
The old mempool imposed hard limits on how many ancestors or descendants a single transaction could have. This created problems. Miners building block templates couldn't always see the optimal grouping of related transactions. RBF replacements could technically meet the old rules while making the overall mempool worse. The CPFP carveout — a special exception added in 2019 to help Lightning Network channels close properly — was a patch on a patch.
Cluster mempool replaces all of this with a single, cleaner model. Instead of tracking individual ancestor/descendant chains, the mempool now organizes transactions into clusters — connected components of related transactions. Each cluster is capped at 64 transactions and 101 kB of virtual size.
Within each cluster, transactions are ordered by feerate chunks — groups of transactions expected to be mined together — rather than by individual transaction feerates. This means miners can now build block templates that more accurately reflect real economic incentives.
Better RBF, No More Carveout
The RBF logic has been tightened significantly. A replacement transaction must now improve the entire mempool feerate diagram, not just beat the fee and feerate of the transaction it's replacing. This eliminates a class of replacements that technically passed the old rules but actually made the mempool worse for miners and users alike.
The CPFP carveout has been removed entirely. With cluster mempool's cleaner accounting, the special exception is no longer necessary — the system handles Lightning channel closes correctly without it.
New Tools for Node Operators
Two new RPCs give operators visibility into the new architecture:
getmempoolclustershows how transactions are grouped and their chunk orderinggetmempoolfeeratediagramdisplays the full mempool feerate distribution
These aren't just developer toys. For anyone running a node — whether a miner optimizing block construction or a business monitoring transaction confirmation — these tools provide a fundamentally better view of mempool economics.
Private Broadcasting: A Quiet but Significant Addition
Buried further in the release notes is a feature that privacy-focused Bitcoiners have wanted for years. A new -privatebroadcast option for the sendrawtransaction RPC routes transaction broadcasts exclusively through Tor or I2P networks, preventing the originator's IP address from being linked to the transaction.
The system uses separate connections per broadcast, making it harder for surveillance nodes to correlate transaction origin with network identity. New RPCs — getprivatebroadcastinfo and abortprivatebroadcast — let operators manage and monitor private broadcasts.
This matters because broadcasting a transaction from your own node has always been a privacy weak point. Even if you use CoinJoin or other on-chain privacy techniques, a well-positioned network observer could still link your transaction to your IP address at the moment of broadcast. Private broadcasting closes that gap for anyone running Tor or I2P.
Other Notable Changes
- Default database cache (
-dbcache) increased from 450 MiB to 1,024 MiB on systems with 4 GB or more of RAM, improving initial block download and sync performance - Embedded ASMap data for IP-to-AS mapping is now available without external files, improving network topology awareness out of the box
- Minimum compiler requirements bumped to Clang 17.0 and GCC 12.1
The Bigger Picture
Cluster mempool has been in development since 2023. It's not a consensus change — it doesn't require a soft fork, and it doesn't change Bitcoin's rules. It changes how nodes organize and prioritize transactions locally. But the downstream effects are meaningful: more efficient block construction, more rational fee markets, and fewer edge cases that sophisticated actors could exploit.
For everyday users, the impact is indirect but real. Better mempool logic means fee estimation improves. Miners building better blocks means less wasted block space. The removal of awkward carveouts means fewer surprises when Lightning channels need to close under adversarial conditions.
The release candidate phase means 31.0 isn't final yet — testing is ongoing, and the stable release is expected in the second half of 2026. But the cluster mempool code has already been merged into the main codebase, and the architecture is unlikely to change significantly before release.
Bitcoin Gate Take
This is the kind of upgrade that doesn't make headlines but quietly makes Bitcoin better. Cluster mempool fixes structural debt that has accumulated over a decade, and the private broadcast feature addresses a real privacy gap without requiring any protocol-level changes. For long-term holders running their own nodes, 31.0 is worth paying attention to — not because it changes what Bitcoin is, but because it improves how well it works.