DeFi
Aave vote backs a borrowing rate limiter for V3
The consultation that closed on November 20 drew about 371,959 weighted votes in favor of LlamaRisk’s proposal. The mechanism would create borrowing headroom that replenishes in stages below governance-set caps.

The vote clears the way for an implementation proposal
The advisory vote on the first stage of LlamaGuard Caps ended on November 20 at 09:28 UTC. All 80 votes recorded on Snapshot supported LlamaRisk’s proposal. It targets Aave V3, where caps already set the maximum amount that can be borrowed for each asset. The consultation clears the way for a proposal to be put to an onchain vote. Activation timing and final parameters still need to be specified.
LlamaRisk’s document says a static cap needs enough space for borrowing to grow until the next adjustment. That space becomes available all at once. The proposed mechanism would control how quickly debt can grow based on the size of each reserve. LlamaRisk presents it as a way to give response teams more time to act when they see unusual activity.
Headroom calculated for each reserve
The system would open headroom above existing debt. As borrowers use it, the mechanism would refill it in stages. A large repayment could also lead it to reduce available space if it exceeds the intended buffer. Aave and its risk stewards would retain control of the maximum cap. Automation would act below that limit to determine how quickly a reserve can take on new borrowing.
LlamaRisk proposes using the Risk Agents infrastructure that Aave governance already owns. Increases would be calculated through Chainlink CRE, with bounds enforced onchain. Aave’s core contracts and existing agents are outside the stated scope. The proposal also provides a dedicated route for integrated institutional borrowers and other approved large borrowers so that ordinary headroom does not block their liquidity needs.
Calibration must account for ordinary borrowing
The size of the buffer and the speed at which it refills matter as much as the limiter itself. Tight settings can delay legitimate borrowing. Broad settings reduce the time gained in an incident. On the forum, IpakSystems presented an independent simulation and noted that the illustrative buffer would have fallen short of fully containing both episodes it examined. The results discussed there come from historical replays.
The first-stage components have entered audits, according to LlamaRisk. The author plans to share more details when the system is ready to deploy. Expansion to supply caps, Aave V4, and risk management across markets belongs to later stages. The November 20 vote covered the pace of borrowing on V3. Other uses would be calibrated using experience from that first step.
