Fix subnet alpha out burn discrepancy

[Dairus01]

Fix SubnetAlphaOut burn discrepancy

Fixes #2274

Problem Description

The SubnetAlphaOut storage value has been incorrectly updated since dTAO launch, causing accounting discrepancies between issued and burned alpha. This breaks the fundamental accounting equation:

total_issuance = subnet_alpha_out + subnet_alpha_in + subnet_alpha_in_provided

burn = subnet_alpha_out - SUM(total_hotkey_alpha)

Burning happens on chain in two ways:

1. The burn extrinsic: Reduces a coldkey's stake without properly reducing subnet_alpha_out
2. Incentive distribution: Increases subnet_alpha_out normally in run_coinbase.rs but doesn't give the alpha to keys owned by the subnet

Both scenarios create a discrepancy between the accounting storage of owned alpha on a subnet (subnet_alpha_out) and the actual stake owned by accounts (sum of total_hotkey_alpha). This has resulted in:

• 6 subnets showing negative burn amounts (at block 6874108)
• Root network accounting off by ~562k TAO
• Incorrect total issuance calculations across the network

Evidence (Block 6874108)

{
  "netuid": 0,
  "diff": "-561931452490774"
},
{
  "netuid":  17,
  "diff":  "-5409469998121"
},
{
  "netuid": 32,
  "diff": "-3529669489799"
},
{
  "netuid": 55,
  "diff": "-2585186709697"
}

Solution

This PR corrects the SubnetAlphaOut updates to ensure proper accounting during:

• Burn operations: Now correctly decreases SubnetAlphaOut when stake is burned
• Stake decrease operations: Ensures SubnetAlphaOut accurately reflects the actual staked alpha

Changes Made

• Modified burn extrinsic to properly update SubnetAlphaOut
• Fixed BalanceOps::decrease_stake to maintain consistency between stake and SubnetAlphaOut
• Added comprehensive regression tests to verify the fixes

Tests Added

Two regression tests have been added in pallets/subtensor/src/tests/regression_issue_2274.rs:

1. test_burn_extrinsic_updates_subnet_alpha_out: Verifies that burning stake correctly reduces SubnetAlphaOut by the burned amount
2. test_balance_ops_decrease_stake_consistency: Ensures that SubnetAlphaOut remains consistent when stake decrease operations are performed, including edge cases

Impact

This fix ensures:

• ✅ Correct burn accounting across all subnets
• ✅ Accurate total issuance calculations
• ✅ Proper tracking of alpha distribution
• ✅ Alpha halvenings will occur at the correct time
• ✅ Max supply of 21M tokens will be respected

Contribution by Gittensor, learn more at https://gittensor.io/

[Copilot]

Pull request overview

This PR fixes a critical accounting bug in subnet alpha tracking that has resulted in negative burn amounts on multiple subnets (including ~562k TAO discrepancy on root network). The issue arose because SubnetAlphaOut was not being properly decremented when alpha was burned through the burn extrinsic and the BalanceOps::decrease_stake interface.

Key Changes:

• Implemented the burn_subnet_alpha function to properly decrement SubnetAlphaOut using saturating_sub
• Fixed BalanceOps::decrease_stake to use the actual decreased amount (not requested amount) when updating SubnetAlphaOut
• Added comprehensive regression tests to verify the fixes

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
pallets/subtensor/src/staking/helpers.rs	Implemented burn_subnet_alpha function to decrement SubnetAlphaOut when alpha is burned (previously was a no-op)
pallets/subtensor/src/lib.rs	Fixed decrease_stake to call decrease_stake_for_hotkey_and_coldkey_on_subnet first and use the actual returned amount when updating SubnetAlphaOut
pallets/subtensor/src/tests/regression_issue_2274.rs	Added two regression tests: one verifying burn extrinsic updates SubnetAlphaOut, another testing edge case of decreasing more stake than available
pallets/subtensor/src/tests/recycle_alpha.rs	Updated existing test assertions to expect SubnetAlphaOut to decrease after burns (previously incorrectly expected no change)
pallets/subtensor/src/tests/mod.rs	Added module declaration for new regression test file

[Dairus01]

@l0r1s @bdmason @thewhaleking @basfroman

Happy Holidays, trust you are doing great, Please I'd love you to review this PR, it helps solves a critical problem

[Dairus01]

This would help solve the subnet alpha out burn discrepancy