Solana Cross-DEX Priority Fee Dataset (2026-04-09 — 2026-05-28)

Fifty days of continuously-collected priority-fee statistics from three Solana DEX aggregators (Jupiter, Raydium, Orca), sampled every 5 minutes between 2026-04-09 and 2026-05-28 UTC. The dataset supports a pre-registered empirical study of cross-venue fee divergence on the Solana network. CONTENTS - 50 daily CSVs (one per UTC day), schema: timestamp + per-venue min/median/mean/max in microlamports per Compute Unit + Solana slot + recent blockhash for provenance. - analyze_fees.py: single-file Python script reproducing every numeric claim in the companion manuscript (Kamat 2026, under review at IEEE Access). Deterministic, no random seeds. - README.md: full data dictionary, reproduction instructions, hypothesis verdict table. - MANIFEST.sha256: SHA-256 hashes of every released artifact. - requirements.txt: Python dependencies. REPRODUCING THE MANUSCRIPT pip install -r requirements.txt python analyze_fees.py Runtime: under 10 seconds on commodity laptop. Outputs results.json, robustness_table.csv, and two figures. SAMPLE - Raw rows: 14,160 - Pre-registered cohort (all three venues nonzero medians): 8,962 observations - Window: 50 continuous days - Cadence: 5 minutes HEADLINE FINDINGS (regenerated from this dataset) - Cross-venue fee divergence exceeds 10x in 90.4% of observations (H1 confirmed) - Raydium quotes the maximum fee in 79.6% of timestamps (H5 confirmed, chi-squared = 8,656) - Divergence correlates with congestion (Spearman rho = 0.555, H3 confirmed) - No hour-of-day periodicity (H2 null, ANOVA p = 0.80) - Divergence does not persist beyond 5 minutes (H4 null) - All three confirmed findings survive four pre-committed robustness subsets. LICENSING - Code (analyze_fees.py): MIT - Data and documentation: CC-BY-4.0 ETHICS AND PROVENANCE Source data: public Solana RPC endpoints and DEX aggregator quote APIs. No private keys, no user identifiers, no personal data. No human-subjects research. Sole author. No external funding. No competing interests. PRE-REGISTRATION The methodology was finalised and time-stamped before the analysis ran on this dataset. Hypotheses, cohort definition, robustness subsets, and significance threshold (Bonferroni-adjusted alpha = 0.002) were all locked prior to results.