Solana’s Geographic Leader Plan Draws New Questions Over Validator Splits

Solana’s proposed geographic leader schedule is drawing a security warning over whether self-reported validator locations could extend maximal extractable value (MEV) opportunities, even as its model estimates faster handoffs.

Solana Improvement Document 0675 (SIMD-0675) would retain stake-weighted leader selection and each validator’s number of leadership windows while changing their order so nearby validators are more likely to lead in sequence. The current schedule assigns four consecutive 200-millisecond slots to each leader window, or 800 milliseconds.

The study modeled 6,250 synthetic networks using 124 observed validator locations across six macroregions. Its latency estimate used routes 50% longer than great-circle distance and signal travel at roughly two-thirds the speed of light. At a 10% stake floor and a seven-window run length, the model estimated an average handoff gain of 29.1 milliseconds. Its cumulative regional-persistence measure increased by 5.4 leader windows, equivalent to about 4.3 seconds. These are simulated estimates, not measurements of live handoffs.

On Oct. 24, a commenter using the handle deanmlittle warned that validators could use self-reported geography to align leader rotations and potentially create longer MEV windows. On Oct. 25, the proposal’s author, posting as wattenhoferSOL, said it should “start with a somewhat short run length.”

On Oct. 26, Umberto Natale, who wrote the Oct. 14 analysis, described a modeled example in which splitting 10% stake between two identities changed the estimated chance of three consecutive leader windows from 7.146% to 7.147%. The scenario used a 10% stake floor and a seven-window run length. Natale said “there should be very little upside in doing sybils.” The estimate does not show that a validator split its stake or settle the broader MEV concern.

The analysis recommends a run length between three and seven, depending on which risk is prioritized; seven is one candidate trade-off, not a universal optimum. The proposal’s design uses validator location information, and a companion proposal, SIMD-0674, would let validators register that data in their vote accounts.

SIMD-0675 remained an open pull request as of Oct. 27, with no mainnet activation date established.