Proposal: Bad Debt Isolation & Repayment Mechanism

[umee-man]

Proposal: Bad Debt Isolation & Repayment Mechanism

Caution

The Problem: In the current x/leverage architecture, when a user's collateral is fully liquidated but borrow balances remain, the remaining debt is flagged as "Bad Debt". However, this debt remains tracked as standard borrow shares. Because interest in Umee is calculated globally via an InterestScalar acting on all borrow shares, bad debt continues to accrue interest indefinitely. This infinitely increases the protocol's deficit and artificially inflates the utilization rate, driving up Borrow APY for healthy users.

1. Executive Summary

This proposal suggests an architectural shift to isolate bad debt. By completely removing uncollateralized debt from the standard borrow tracking system and moving it to a static, global ledger, the protocol can freeze the debt size (stopping interest accrual) while continuing to pay it down programmatically from newly accrued protocol reserves.

2. Proposed Architecture

The solution requires two primary state transitions: isolating the debt to freeze its value, and updating the global sweep mechanism to pay it down efficiently.

A. The Isolation Mechanism (Freezing the Debt)

Currently, the liquidation hook postLiquidate flags an address|denom when collateral hits zero.

Proposed Change:
Instead of just flagging the address, the protocol should execute a debt write-off during the liquidation state transition:

1. Calculate Final Debt: Convert the user's remaining borrow shares into a static base token amount (e.g., 100 USDC).
2. Erase the Position: Delete the user's borrow position entirely so they no longer hold borrow shares.
3. Decrease Total Borrowed: Subtract the static amount from the leverage module's TotalBorrowed state. This immediately fixes the utilization rate calculation, ensuring healthy borrowers are not penalized with high interest rates caused by dead liquidity.
4. Move to Isolated State: Add the static amount to a new global variable: IsolatedBadDebt[denom].

B. Modified Reserve Sweep Procedure (Automatic Repayment)

Currently, the SweepBadDebts function (called in EndBlocker) iterates over all flagged address|denom pairs across the network, attempting to repay each individually. This is highly inefficient.

Proposed Change:
The SweepBadDebts function should be refactored to operate globally rather than individually:

1. In EndBlocker, the protocol checks if IsolatedBadDebt[denom] > 0.
2. It checks if the module has accumulated Reserves[denom] > 0 from recent interest accruals.
3. If both are true, it calculates the repayment amount: Repayment = min(IsolatedBadDebt, Reserves).
4. It subtracts Repayment from both the IsolatedBadDebt tracker and the Reserves pool in a single O(1) state operation.

3. Key Benefits

Benefit Area	Description
Protocol Solvency	Prevents bad debt from growing exponentially via compound interest, allowing the protocol's reserves to actually catch up and eliminate the deficit.
Market Fairness	Removing dead debt from TotalBorrowed restores accurate utilization rates. Healthy borrowers will no longer pay artificially inflated APYs caused by bad actors.
Performance	Replacing an O(N) network-wide iteration over individual bad debt addresses with a single O(1) global math operation saves significant computation overhead in the EndBlocker.
User Experience	Liquidated users no longer see an infinitely growing negative balance that they can never realistically repay.

4. Implementation Steps

1. Introduce a new KVStore prefix for IsolatedBadDebt structured by denom.
2. Update x/leverage/keeper/liquidate.go -> postLiquidate() to wipe the borrow position and map it to IsolatedBadDebt.
3. Update x/leverage/keeper/iter.go -> SweepBadDebts() to utilize the global O(1) repayment logic.
4. Ensure ReserveFactor continues accruing from healthy TotalBorrowed balances to fund the repayment.

[umee-man]

Adding a few clarifications to avoid potential misunderstandings and strengthen the proposal.

First, this mechanism should not be interpreted as a “write-off” in the economic sense. The loss is not eliminated — it is reclassified. The goal is to remove unrecoverable debt from the active borrow-share accounting system and move it into a protocol-level bad debt ledger.

Once a position has no collateral and no recovery path through standard liquidation, keeping it inside the interest-bearing system leads to artificial growth of bad debt and distorted utilization metrics. Isolating it ensures that healthy borrowers are not penalized by inflated APY caused by dead liquidity.

It’s also important to explicitly state that the protocol’s solvency model remains consistent. The loss is still accounted for — just no longer within TotalBorrowed, but in IsolatedBadDebt[denom], representing a deficit to be repaid from reserves or other governance-defined mechanisms.

Another important aspect is migration of existing bad debt. A one-time process would be required to:
– calculate current residual debt for all flagged accounts
– remove their borrow shares
– adjust TotalBorrowed accordingly
– move the amounts into IsolatedBadDebt[denom]

Finally, adding some safeguards could improve robustness:
– optional per-block or per-denom caps for repayments
– clear event emission for isolation and repayment actions
– explicit linkage between reserve accumulation and bad debt reduction

Overall, the key idea is that bad debt without collateral should not remain inside the live interest-bearing system, as it creates a structural imbalance and prevents the protocol from recovering effectively.

[robert-zaremba]

Indeed, the Bad Debt is not correctly handled - it's a bug to accure interest on a fully liquidated position with a bad debt.

The proposal makes sense and solves the bug.