Solana study backs a geo-aware leader schedule — and is unusually frank about its risks

A technical study published on Solana's own site on October 14 evaluates SIMD-0675, a proposal to reorder Solana's leader schedule by geography so that consecutive block producers sit physically closer together, shaving propagation delay off each handoff.

The mechanism leaves the existing stake-weighted random selection untouched. It takes that schedule as input and only reorders it, so every validator keeps exactly the same number of leader windows per epoch — currently windows of four consecutive slots each — and only their sequence changes. Two tunable parameters drive the reordering: STAKE_FLOOR (σ), which sets how large a radius counts as "nearby" for each location, and RUN_LENGTH (γ), the size of the contiguous blocks of geographically compatible leaders.

The analysis tests these on 6,250 synthetic Monte Carlo networks built from Solana's current 124 validator locations, grouped into six macroregions, with stake and nodes redistributed at varying concentrations. Latency is estimated purely from distance, assuming fiber at about two-thirds light speed over paths 50% longer than the great-circle route.

The headline recommendation: σ around 10% and γ between 3 and 7, yielding roughly a 29 ms mean handoff gain over random at γ = 7. The study is explicit that these conclusions hold only for today's geography and should be revisited if stake shifts between regions or new locations open up.

Key facts (all from the Solana study, Oct 14, 2026):

• At σ = 10%, mean handoff gain runs from 15.8 ms at γ = 2 to 33.9 ms at γ = 50.
• Gains saturate fast: γ = 3 already captures 65% of the γ = 50 gain, γ = 5 gets 79%, γ = 10 gets 91%.
• σ is capped near 10% because Asia holds about 10.4% of stake today; above 10%, Asia loses its local bins.
• Geographic persistence worsens with larger γ: at γ = 50, excess same-region persistence falls from ~61 leader windows (σ = 2%) to ~23 (σ = 20%).
• A companion proposal, SIMD-0674, would require validators to register their location.

The real-world read. This is Solana's own research operation analyzing a Solana proposal — a primary source, but an interested one, so treat the framing as the project making its own case. To its credit, the study does not oversell: it states plainly that benefits plateau above γ ≈ 10, that latency is bought with longer stretches of one region holding the network (raising exposure to regional outages, censorship or partitions), and that it introduces a fake-location attack surface — validators lying about where they run to game the schedule. SIMD-0674's location registration is what makes that spoofing possible in the first place, and higher σ is partly a defense against it. The gains are real but modest and geography-dependent, and nothing here is shipped: this is a parameter study on a draft proposal, not a deployed change. There is no independent analysis to corroborate the latency figures.

Opinion, and whose. The study's authors argue the handoff improvement eases "pressure for stake concentration in macroregions" and that at σ = 10% the simulated rise in MEV opportunity from consecutive same-region leaders is limited. Those are the authors' modeled conclusions, conditional on current geography — not measured mainnet outcomes.

Sources. Solana News, "Geo-Aware Leader Schedule," Oct 14, 2026 (solana.com) — primary source for the SIMD-0675 proposal, parameters, simulation methodology and all figures. Published by the project itself; not independent third-party research, and not corroborated elsewhere here.

Not financial advice.