Compact block relay in knots using the extra pool

Hi bitcoin developers,

Compact block relay has been one of the concerns with divergence in mempool policies. I tested it in knots with a pruned and non-listening node. The results are interesting and details are shared in this post: https://uncensoredtech.substack.com/p/compact-block-relay-with-extra-pool

I used this config for the node:

debug=cmpctblock feefilter=0 blockreconstructionextratxnsize=300

It was running for ~10 hours and had 8 full relay outbound connections in which 2 were bitcoin core nodes (v29.1 and v30.0) that relayed sub 1 sat/vB transactions. 90% blocks had no issues with reconstructions because of less transactions being requested.

Example: Block 919005

2025-10-14T09:14:50Z Saw new header hash=00000000000000000000f76ea0d9598e36231496c732a95fab75fc91c8d5bcf2 height=919005 2025-10-14T09:14:50Z [cmpctblock] Initialized PartiallyDownloadedBlock for block 00000000000000000000f76ea0d9598e36231496c732a95fab75fc91c8d5bcf2 using a cmpctblock of size 28617 2025-10-14T09:14:51Z [cmpctblock] Successfully reconstructed block 00000000000000000000f76ea0d9598e36231496c732a95fab75fc91c8d5bcf2 with 1 txn prefilled, 4440 txn from mempool (incl at least 3845 from extra pool) and 235 txn requested 2025-10-14T09:14:51Z UpdateTip: new best=00000000000000000000f76ea0d9598e36231496c732a95fab75fc91c8d5bcf2 height=919005 version=0x200c2000 log2_work=95.878521 tx=1256146478

Related pull requests: https://github.com/bitcoinknots/bitcoin/pull/217 https://github.com/bitcoinknots/bitcoin/pull/218

Feel free to share your feedback and other ideas that can improve compact block relay.

/dev/fd0 floppy disk guy