Allocate TAO emissions across the top 64 subnets

[gzaentz]

Summary

• Allocate 75% of block TAO emissions to subnet ranks 1-32.
• Allocate 25% to ranks 33-64.
• Allocate no TAO emissions to ranks 65 and below.
• Preserve score proportionality within each tier.
• Resolve equal-score boundary ties by ascending netuid.
• Give the eligible set 100% of emissions when 32 or fewer subnets have positive scores.
• Document the behavior and add focused boundary and proportionality tests.

Rationale

The current top-32 cutoff concentrates all TAO emissions among today's highest-ranked subnets. This strongly rewards proven performers, but it also creates an incumbent advantage: a credible subnet immediately below rank 32 receives no TAO emission even when it may be close to product-market fit or capable of becoming a future top performer.

This proposal introduces a deliberate exploration and exploitation balance. The top 32 retain a dominant 75% share, so most capital remains concentrated on subnets that have demonstrated demand and performance. The next 32 collectively receive 25%, giving the competitive middle enough support to improve infrastructure, attract participants, and prove whether they can challenge current leaders.

This is not an equal subsidy. Rewards remain proportional to performance within each tier, only the top 64 positive scores qualify, and a subnet must improve its ranking to enter the larger 75% pool. The intended benefit is greater network option value: preserve strong support for current winners while increasing exposure to potential breakout subnets and keeping the top tier contestable.

User and developer impact

Subnet operators ranked 33-64 gain a bounded TAO emission path. Top-32 subnets continue to receive most emissions, while ranks 65 and below remain ineligible. The ranking is deterministic, including equal-score ties, and no block emission is stranded when the second tier is empty.

Validation

• 3 focused ranked-tier tests passed.
• All 14 subnet-emission tests passed.
• All 78 coinbase tests passed.
• All 1,423 pallet-subtensor all-feature tests passed, with 9 ignored.
• All remaining Rust workspace tests and doctests passed after isolating the known failures below.
• Separate eco-tests: 56 passed.
• Documentation preview tests: 56 passed.
• Python SDK pytest: 980 passed and 1 skipped in the sandbox; the localhost bridge tests passed when rerun with socket permission.
• Rust formatting and git diff --check passed.

Known unrelated repository failures

The complete repository suite is not currently green. Four runtime proxy-filter tests and the Python SDK codegen coverage check fail because the existing call inventory omits sudo_set_emission_bar_quantile and sudo_set_emission_gate_exponent. Ruff formatting also reports two unrelated Python SDK files. None of these failures touch the three files changed by this PR.

[vercel]

@gzaentz is attempting to deploy a commit to the RaoFoundation Team on Vercel.

A member of the Team first needs to authorize it.