What happened

IOG merged PR #1057 for the Leios dozen-devnet. The change adds colour-tagged transaction generators and one mempool monitor per node, allowing each node to report how much of its mempool comes from its own block producer versus transactions that arrived from elsewhere.

A 5 MB mempool cap is included so that nodes must hold a selection of transactions, making fragmentation observable across the 12-node cluster.

Why it matters

How transaction load is distributed across node mempools is relevant to whether Leios can support higher transaction throughput. This tooling is designed to measure that behaviour directly rather than infer it from a single cluster-wide view.

For stake pool operators and builders following Leios-era capacity work, the useful signal is the resulting fragmentation and throughput evidence, not the merge alone.

What to watch next

Watch for devnet results showing how much transaction load remains local to a node's block producer, how much arrives from elsewhere, and how those measurements relate to throughput. The supplied record also notes a dashboard that breaks transaction-submission time into idle, load, and submission fractions.

What to watch

Published dozen-devnet measurements of fragmentation and throughput from the new per-node monitors.

Sources and limits

Upstream references and independent checks

Digest dated 2026-08-29 · upstream model claude-sonnet-4-6. Direct links are matched to all 3 upstream source IDs.

  1. 1
    Merge pull request #1057 from input-output-hk/ch1bo/mempool-fragmenta…Direct upstream source · 1ded0846ca23fc26f8da349a3cbacf2a65984724
  2. 2
    devnets: colour the generators and observe each mempoolDirect upstream source · 9a9335dedcfdb5c529ed8c0657414fd080ebe600
  3. 3
    dozen-devnet: cap the mempool at 5 MB for fragmentation testsDirect upstream source · 93ea66831d0ddeeefe7e551a5a790a880c1429bb

This Research brief was generated by Terra from a dated upstream research digest. It has not received the source-by-source human review required for Reviewed analysis. Material limit: This is single-source, medium-confidence devnet instrumentation from a merged change; it is not a protocol change, public testnet, mainnet deployment, or throughput result.