Inside the Block: Validator Ordering Lab

u64se-block-order (lab/block-order) is a dedicated empirical research and measurement module inside u64se-seu-solana designed to evaluate how Solana validator schedulers order transactions into blocks.

• Network: Solana Mainnet-Beta (Post-Alpenglow Upgrade, Epoch 1073)
• Measurement Date: November 1–2, 2026 (Slots 463,931,745 to 464,374,627)
• Validator Diversity: 10 distinct Mainnet validators across 4 distinct client software versions:
  • Agave 4.3.2
  • Agave 4.4.0
  • Agave 4.4.0-rc.1
  • Frankendancer / Firedancer 26.10.5
• Total Mainnet Spend: 0.001017 SOL (across 20 live Mainnet probe transactions, strictly bounded by an atomic SpendGuard limit of 0.0015 SOL)
• Formal Guarantee: Backed by Theorems 13, 14, and 15 in verify_invariants.smt2.
• Interactive Telemetry Visualizer: seu-solana.rag.engineering/block-ordering.html

Executive Summary & Core Findings

Rather than treating Solana block production as a naive First-In-First-Out (FIFO) queue or a one-dimensional priority fee auction, this module injects deterministic, self-identifying probe transactions carrying a canonical 64-bit StateExecutionUnit register word (u64se:probe:0x...) directly into Mainnet leader windows.

Three Key Results:

1. Intra-Block Inversion via Requested CU Limit (Kendall’s $\tau = -1.000$):
   At flat 5,000 lamports base fee and zero priority tip, reducing requested compute budget strictly elevates scheduler priority score. On Frankendancer v26.10.5 (Slot 463935349), a probe requesting 30,000 CU jumped 51 positions ahead of a prior probe requesting 50,000 CU. On Agave 4.3.2 (Slot 464360092), all concurrent probes sorted in exact inverse order of arrival.
2. Sub-2,000 CU Infiltration Ahead of Jito MEV Bundles:
   In Slot 464374627 (Agave 4.3.2), our seu-checkpoint preset (1,900..2,500 CU) captured 4 consecutive block slots (indices 27, 28, 29, 30 / 375; top 7.2% of block) ahead of 13 out of 14 Jito MEV bundles that paid substantial tips but were appended to later PoH ticks.
3. 100x Scheduler Leverage for < 1,900 CU Programs:
   Setting a tight ComputeUnitLimit(2,000) yields a 100x higher validator reward density ($2.50\text{ L/CU}$ vs $0.025\text{ L/CU}$) at zero priority fee compared to standard 200,000 CU defaults.