[ARC] Aave ETHPoW Fork Risk Mitigation Plan

[Primoz]

Hey all,

Block Analitica recently received a grant from the Aave Grants DAO for building the Aave Analytics Dashboard. The dashboard is currently in the second stage of a three-part development plan. More information is available on the Aave Grants DAO website. Block Analitica believes that the Aave Analytics Dashboard can provide useful tools and data that may help the decision-making process of Aave governance. This process can be described as “Computer Aided-Governance”. The dashboard is intended to be used to make better-informed decisions by enabling a high-level view of the Aave v2 protocol on Ethereum.

This post highlights certain risks that Block Analitica has identified as it relates to Aave v2 on Ethereum and the upcoming Proof of Stake (PoS) merge. These risks are derived from and illustrated by data from the Aave Analytics Dashboard. Finally, this post includes a potential mitigation strategy and proposed parameter changes to the ETH market on Aave v2 on Ethereum.

Background

@monet-supply from our team already made a great overview of Ethereum merge risks for MakerDAO. After some discussion, we realized that lending markets such as Aave and Compound face much greater risks. Most tokens, excluding ETH, will likely be worthless on an ETHPoW chain. Hence, one strategy users may employ to maximize their cryptoasset holdings, will likely be to borrow as much ETH as possible, collateralized mainly by stablecoins or other tokens. In addition, we should probably also observe flows from stETH holders to ETH, since stETH will likely also be worthless on an ETHPoW chain.

Supported Data

Speculative strategies related to the PoS merge and the potential ETHPoW fork will likely have implications for Aave, particularly because Aave enables ETH to be borrowed from stETH. stETH collateralised ETH borrowing has become a popular strategy and has increased ETH market utilization to a level of 62% at the time of writing. A majority of ETH borrowed is currently collateralised by stETH.

As illustrated in the chart below, large part of ETH that is currently borrowed is collateralised by stETH (approximately $907 million in stETH collateral is backing ETH, represented as the dollar-denominated ETH amount borrowed from stETH divided by the LTV ratio). The stETH depeg situation recently improved as a result of the merge date prediction which, in turn, led to the stETH/ETH recursive position strategy to regain popularity and grow in size.

Luckily, stETH/ETH positions maintain a high buffer. A discount of stETH to ETH would need to reach around 15% to start experiencing larger liquidations. Further information is provided in the chart below. Note that the chart assumes a partial liquidation scenario in chunks of a maximum of $5 million, until the position health ratio is improved, which is based on empirical data of Aave liquidations.

Examining the chart of collateral backing borrowed ETH in more detail (see chart below), reveals another potential issue that may arise: ETH is increasingly being borrowed and collateralized by other, non-stETH assets, such as USDC, WBTC, or even ETH itself. The value of non-stETH collateral for ETH borrowing has increased from $120m to $390m in the last 10 days. This trend suggests that our hypothesis about the ETH market becoming increasingly utilised up until the merge is correct.

By studying individual Aave wallets and strategies through the Aave Analytics Dashboard we noticed an unusual behaviour of users building ETH/ETH recursive positions. There are at least two larger wallets that we have identified. The screenshot below includes an example of one such wallet. This strategy enables ETH holders to maximize their ETH that will be redeemable on an Aave ETHPoW fork. However, this strategy might not work since ETHPoW intends to freeze aWETH contract.

Risks

So what could go wrong? We have identified three main risks related to the ETH market becoming increasingly utilized:

High ETH utilization potentially makes liquidations harder or impossible

In a situation where the ETH market becomes heavily utilized and markets start experiencing high volatility as a result of the merge event, liquidations of regular ETH long/stablecoin short positions might not be possible. This will be due to the fact that liquidators will not have access to ETH as collateral since the majority of ETH will be borrowed. This could in turn lead to some positions becoming uncollateralized. Note that because people are now building stablecoin long/ETH short positions, a price increase in ETH could also start liquidating stablecoin collateral, which will be less of an issue, but still something to be aware of.

High ETH utilization increases the ETH rate to a level where stETH/ETH positions are making negative APY

Once the ETH borrow rate reaches 5%, which happens shortly after 70% utilization rate (we are at 63% right now), stETH/ETH positions start becoming unprofitable. Currently, borrowers on Aave are not leveraged maximally due to depeg risks. As a result, it is possible that some positions will have their APY negative much earlier. This will cause users to unwind their positions up until the ETH borrow rate reverts to a stable level where the APY becomes tolerable. This means that we would see a lot of stETH to ETH redemptions and in turn a downward push on the stETH price, which will already be under pressure due to regular stETH holders switching to ETH to gain upside on ETHPoW work. All of this can lead to a downward price spiral of stETH which can create cascading liquidations at Aave. Hopefully people start bidding stETH at a not-too-large discount, which is more likely now that the merge timeline is clearer.

Already high ETH utilization causes regular ETH suppliers to start withdrawing their ETH

As a result of the uncertainty and risks related to the ETHPoW fork and PoS merge in general, and ETH utilization on Aave in particular, current liquidity providers may become increasingly worried about their ETH on Aave, and in turn, may withdraw ETH on the supply side. The utilization rate could increase unrelated to ETH borrowers,. Also, if ETH price drops then ETH long/stablecoin short positions will likely need to be deleveraged by selling supplied ETH. In such a scenario, the ETH utilization would increase even further. One of the ETHPoW Twitter accounts is also already calling everyone to withdraw assets from lending platforms, which could create additional pressures on high ETH market utilization.

Potential Mitigations

Based on our views of the various risks related to the PoS merge and potential ETHPoW fork, we recently proposed changes to the ETH Interest rate model at Compound, including setting ETH borrow cap to 100.000. To our knowledge, Aave v2 does not have the capability to limit ETH borrowing to a specific number, but it has the option to disable borrowing. As much as this sounds like an overreaction, we believe that Aave should freeze ETH borrowing in the interim period leading up to the merge.

However, this action alone might not be enough. As previously illustrated, a high utilization rate of the ETH market could in fact increase unrelated to ETH borrow dynamics. Therefore, a more aggressive interest rate increase for the ETH market is proposed. However, this will create certain tradeoffs. A higher ETH borrow rate would force stETH/ETH recursive positions to be closed, potentially worsening the stETH depeg situation. On the other hand, the discount would need to be severe (15%+) to cause liquidations in the tens of millions of USD value.

In our view we believe that a more aggressive rate curve should be implemented, since the potential ETH borrow freeze might not be enough to prevent high market utilization. Ideally, the new curve should not affect the current stETH/ETH positions and cause massive unwinding. Therefore, a steeper slope is only proposed for the part of the ETH rate curve after the kink. Instead of a 103% rate at 100% utilization, we propose to set it to 1,000%.

In conclusion, the following parameter changes to the Aave Ethereum v2 ETH market are proposed in the interim period leading up to the merge:

1) Freeze ETH borrowing.

2) Increase the variable borrow APR at 100% utilization from 103% to 1,000%.

AAVE Payment

If this proposal is adopted, a one-time payment of 60 AAVE will be transferred from the Aave governance to the MakerDAO pause proxy. This is intended to compensate for Block Analitica’s research and development costs in connection with this proposal. BA provides risk management consulting services to MakerDAO as the Maker risk core unit.

Transfer 60 AAVE to 0xBE8E3e3618f7474F8cB1d074A26afFef007E98FB

[monet-supply]

I have deployed a factory pool on Curve with ETH/aWETH. This could offer a second line of defense against insolvency risk in the event that the aWETH market becomes 100% utilized in the lead up to the merge; liquidators or users seeking to unwind positions would be able to swap aWETH for ETH directly using this pool’s liquidity.

This would incur trading fees as well as potentially high slippage (depending on total deposited liquidity), so hopefully the above proposed parameter changes will be sufficient to prevent 100% utilization in the first place.

The pool can be found here: Curve.fi

Note that the pool has not yet been seeded with liquidity, so swaps won’t work until this has been added any initial liquidity providers should be sure to add ETH and aWETH in equal amounts.

As an alternative, it may be possible to set up a concentrated liquidity pool on Uniswap v3 using a constant balance wrapper for aWETH such as this token used by Alchemix (rebasing tokens such as aTokens are not compatible with Uniswap v3). An example of a Uniswap pool for lending market tokens can be found here (includes ETH and Compound cETH).

[Pauljlei]

Thanks, @primoz, for the proposal.

Gauntlet is supportive of pausing ETH borrowing, specifically, calling disableBorrowingOnReserve to disable ETH borrowing. However, it would be prudent to break up this proposal into separate parts and perform more analysis before increasing Aave’s ETH borrow APR at 100% utilization.

The efficacy of the interest rate change depends on the expected yield of fork farming (expected ETHW price / ETH price) and the duration of ETH borrowing (time between borrowing ETH and selling ETHW to USD on a centralized exchange). The effect of the PoW fork is similar to the effect of a positive Ampleforth rebase, where both create profit opportunities at a scheduled time. We conducted a similar analysis of the relationship between AMPL rebase and AMPL interest rate. Similar to the positive AMPL rebase, if ETHW retains non-negligible value, raising the interest rate cannot completely eliminate 100% pool utilization since the borrow duration can be short (within minutes), which suggests that while increasing the ETH interest rate can reduce the time of high utilization, it won’t completely solve the 100% ETH utilization issue. In addition, the OP mentioned that increasing the ETH interest rate will force deleverage stETH/ETH recursive positions and worsen the stETH / ETH price deviation. We’d add that there’s a non-zero risk of liquidation cascades from people having to deleverage under high ETH utilization, creating more market uncertainties.

The tradeoffs need to be carefully analyzed before increasing the max interest rate. We’d encourage the OP to provide more justification on the below with regards to the proposed interest rate curve changes:

• How much additional stETH sell pressure will this create due to stETH/ETH recursive position deleveraging? What are the projected stETH insolvencies caused by the stETH / ETH price deviation? At the time of writing, Aave astETH contract is the largest stETH holder and holds 21% of stETH. The ETH pool utilization is 63%.
• How many potential liquidation failures caused by high ETH pool utilization can this prevent? What are the projected insolvencies prevented by avoiding liquidation failures?

Understanding these tradeoffs is important to make an informed decision on the proposed interest rate curve changes. Gauntlet would suggest splitting this proposal into two proposals (disable ETH borrowing vs. interest rate curve changes) so the community can vote on these two changes independently.

Given the time sensitivity, it is better to start the governance process sooner rather than later. A Snapshot vote for disabling ETH borrow should be published tomorrow. In addition, given the complexities of on-chain governance, Gauntlet is also happy to support publishing the AIP for disabling ETH borrowing.

As for the interest rate changes, the OP can propose those changes if needed. We’d encourage more justification, as outlined above, before doing so, given the added risks of changing the interest rate curve.

[Primoz]

Hey @Pauljlei thanks for feedback. We agree to split the proposal in two and start the ETH borrow pause snapshot as early as possible.

As explained in our analysis, the higher rate curve for ETH has many tradeoffs, specifically because of stETH/ETH recursive leveraged positions. We are aware that a higher rate might put additional pressures on these positions to unwind and this is why we proposed to increase the rate only after kink.

Let’s assume ETH borrow gets disabled but ETH supply withdrawals start to occur for some reason and utilization starts increasing towards 100%. Speculators borrowing ETH to gain yield on ETHPoW fork are potentially prepared to carry up to 3% cumulative cost, assuming the forked ETHPoW may be trading at 0.03-0.04 ETH (currently on Poloniex). The same max cost would be then also carried by stETH/ETH recursive positions. These users are probably counting on successful merge and the current 3% stETH discount to improve. Assuming this, we believe the negative APY impact for stETH/ETH positions is quite limited.

But, we could also have a situation where ETH market utilization increases over 70% and stays there for a longer duration of time. This would be unrelated of ETHPoW positions, but because of some longer lasting run on ETH deposits. The cumulative cost for stETH/ETH positions could reach over 3% and here some users might start to break and unwind. The question then becomes how much stETH selling could we witness until the utilization improves back below 70% where APY becomes positive again. There is around $800m ETH borrowed right now and at a potential 80% utilization it means we’d be seeing around 65k stETH sold to repay ETH borrow (about 33% of the curve stETH/ETH pool) in order to get back below 70% level. This pushes the price of stETH from 0.970 to 0.955 ETH, which has almost no impact on cascading liquidations. In the worst case scenario of 100% utilization and current borrow market size, we’d need to see about 160k stETH sold and ETH repaid (again to get back to 70% ETH utilization) which pushes stETH price to 0.91. As much as 10% depeg sounds devastating to stETH/ETH positions, most of them are well collateralizated (see chart below and note that this chart doesn’t assume partial liquidations so liquidated amounts would be much smaller).

Another concern of high ETH utilization is that in an extreme case where large amount of stETH liquidity is pulled from Curve or Balancer, the liquidators wouldn’t be able to source enough on-chain liquidity to liquidate stETH/ETH recursive positions. Therefore it might be better to rely on users unwinding prior to such event.

TL;DR is that we see risk of high ETH market utilization and its impacts higher than the risk of having stETH/ETH positions unwound by users due to potentially high negative APY caused by higher rate, where impact on cascading liquidations is limited. This is why a higher rate at 100% utilization was proposed. However we believe ETH borrow pause alone should hopefully be enough to prevent high utilization of ETH market. If there is an evidence of ETH supply dropping, we would advise to increase rate at 100% immediately.

[Pauljlei]

Snapshot vote to pause ETH borrowing can be found below. Voting begins on Tuesday, August 30th.

https://snapshot.org/#/aave.eth/proposal/0xa121311c67b7a5bbe5b8b5fe1911663a0ab94ed339a6a4b0e1b9443f670a0e97

[MarcZeller]

Hello,

first, thanks for this very detailed proposal.

While the pausing of ETH borrowing is the safest bet, I have to confess I’m not very keen on altering the free market.

This is a typical game theory scenario with an optimal strategy of every liquidity protocol hitting the pause button and mitigating risk.

Unfortunately, a single liquidity protocol not following the safest bet will have an unfair advantage with a very high supply rate and the opportunity to drain significant deposits (If u earn 3% on your aETH but can make 80% on another liquidity protocol what do you do?) with the risk of these deposits not coming back after the merge.

secondly, I don’t think the interest rate strategies should be used for short-term reaction, the stability of the protocol and the trust of our users relies on the fact that DeFi is at least partly, “predictable”.
Altering the rules regularly for “the greater good” is a slippery slope and current ETH depositors/borrowers might frown at the Aave protocol changing the “deal” mid-game drastically.

evolution of Risk parameters to follow typical market conditions is expected, but 10x the maxRate is very drastic.

Lastly, in every scenario, I’m in favor of the payment of the MakerDAO pause proxy.

to sumup:
I think user safety should be the top priority, and while not loving the idea, I’m supportive of the ETH borrow temporary pause.
That said, I’m expressing my concerns regarding any alteration of the interestRate strategy.

[Pauljlei]

On-chain proposal to pause ETH borrowing is now live for voting:

Aave - Open Source Liquidity Protocol

[eboado]

Complicated topic, but after assessing pros/cons of keeping ETH borrowing enabled I’m in favor of stopping borrowing.

Yes, it could create some asymmetry with other protocols keeping it enabled, but let’s be honest, those protocols are doing something really irresponsible given the realistic potential risk of full utilization of ETH, and not having collateral available for liquidation.

In a second point, as an AAVE holder and community member I find 2 aspects of this proposal a bit concerning:

1. Needless to say, the analysis of the OP is of high quality, and really nice to see other entities participating on Aave and trying to help. That being said, we are setting a really dangerous precedent with this and also some other proposals: somebody does something valuable for the community → asks for a sum of funds directly on the analysis itself.
   This is not the first case, with for example ARC: Aave V3 Retroactive Funding just approved (even if completely different order of magnitude).
   My point is that we are normalizing this way of acting, and it is not normal. And it is especially bad when linking even with on-chain proposals the action itself (stopping borrowing) and the release of the funds to a 3rd party.
   In this case, I’m completely in favor of the sum requested, but this should be done through the Aave Grants DAO. So it is difficult to understand that even the proposer (Gauntlet) includes that on both Snapshot and on-chain. It should just not be included at all.

2. As an AAVE holder, I would like to understand why Gauntlet (represented here by @Pauljlei ), the party directly engaged with the Aave community and getting paid for it, takes so much time to not only do the risk analysis pre-merge but even to answer to a completely un-engaged party like Block Analitica on their analysis.
   I have supported Gauntlet in the past, (and worked with them on multiple topics around Aave) but I feel quite unacceptable that:

   • We are 10 days away from an event as potentially disruptive as the Merge, and only 1 week ago the main party engaged for risk purposes shows involvement in the topic.
   • Some non-engaged party needs to appear to give an assessment, and it takes 7 days to answer them.

   The community is paying a fair sum of funds for the risk-related services, so I think it is important to understand which was the problem here.

[sakulstra]

I think this proposal wont work out as expected in it’s current form and therefore i’m against.
The problem that the powFork creates is not borrowing, but wETH liquidity leaving the protocol, which the proposal doesn’t stop.

When stopping borrowing on the pool we only break the possibility of things normalizing after the fork.
There’s a very high incentive for ppl withdrawing already which is only increased by disabling borrowing.
So eventually utilization will still reach 100%.

I first thought that scenario was unlikely, but when checking the aWETH holding:

Aave interest bearing WETH (aWETH) Token Tracker | Etherscan

you can see that in the top 5 holdings (ignoring the first), it’s addresses without debt and addresses with stable wETH debt, meaning when this proposal is getting executed there’s a high incentive to move to e.g. compound for a 1000% yield (as on aave they can’t even reach 100% due to disabled borrowing) so liquidity will still dry up, just while also losing TVL.

Disclaimer: I’m not experienced in risk topics at all, but this feels rushed.

[monet-supply]

@sakulstra

there’s a high incentive to move to e.g. compound for a 1000% yield

Compound is implementing a borrow cap, so it is extremely unlikely that this maximum 1000% yield will be reached. We would have proposed similar on Aave if this feature was available in v2, but in this case pausing borrowing was the best available risk lever to avoid 100% utilization.

@sakulstra

The problem that the powFork creates is not borrowing, but wETH liquidity leaving the protocol, which the proposal doesn’t stop.

As a non-custodial protocol, it is generally not acceptable to prevent users from withdrawing their funds. However, I think the risk of massive withdrawals is somewhat lower because it would take a large number of uncoordinated existing aWETH suppliers choosing to withdraw to force the market to 100% utilization, while even a single well capitalized user could borrow all of the remaining WETH from the protocol.

[sakulstra]

I think the risk of massive withdrawals is somewhat lower because it would take a large number of uncoordinated existing aWETH suppliers choosing to withdraw

My main argument is that I don’t believe this assumption to be true.

GnosisSafeProxy | Address 0xeaa7723633cf598e872d611f5ec50a45b65cbc72 | Etherscan 3.9% of total supply, 0 debt
InstaAccountV2 | Address 0x0cd420966fce68bad79cd82c9a360636cf898c1e | Etherscan 2.94% of total supply, dai debt - might unwind for better yield or powETH
Strategy | Address 0xf554e20a695ea338a126a3a7b5870f1f67253b01 | Etherscan 2.1% of total supply, 0 debt
Address 0x68030330e8158be3fa5b3ec3c94bf07e42824b9b | Etherscan 1.8% of total supply, eth stable debt - currently losing money and likely to unwind when utilization doesn’t go to 100%
GenericAave | Address 0xe26684de8d5a40a35c3fc15c9328c7f3a523759b | Etherscan 1.8% of total supply, 0 debt

I very non-scientifically only checked the first few(8) addresses on etherscan.
From the first 8 addresses 4 can simply withdraw without even having to unwind anything - so it’s more than likely i guess that 15+% will of supply will disappear.
From looking at it it seems like the only rational things for them would be to withdraw before the merge - if it’s due to better rates on other protocols or due to extracting value from powETH (likely both).

edit: some more addresses
0xa688bc5e676325cc5fc891ac48fe442f6298a432 1,7%, low non-eth debt
0x32e2665c8d696726c73ce28acee310bfac54db85 1.6%, low non-eth debt
0xd22d32a43758186b97cb77dca21d8dfb6515d197 1.6%, low non-eth debt
0xef001a0d62d43ccab7ac9c461f538e707a9edbf2 1.5%, low nen-eth debt

[monet-supply]

@MarcZeller

Unfortunately, a single liquidity protocol not following the safest bet will have an unfair advantage with a very high supply rate and the opportunity to drain significant deposits (If u earn 3% on your aETH but can make 80% on another liquidity protocol what do you do?) with the risk of these deposits not coming back after the merge.

It’s tough to know how users will react to these potential changes, and there could be some chance that more risk tolerant protocols would pick up users. But I think in general users’ hierarchy of needs/wants is:

• minimize insolvency risk / possibility of losing their investment principal
• ensure liquidity / ability to move or withdraw funds as needed
• maximize yield

So Aave taking action to limit liquidity risk could be a positive, demonstrating to users that they can trust Aave governance to respond to changing market conditions. Aave’s relationship with users is based on predictability, but maintaining second order characteristics like protocol safety and liquidity could be more important vs specific parameter settings.

[monet-supply]

How do you see continuing to allow borrowing WETH improves the risk/liquidity situation? Although it may not completely solve the liquidity issue, it seems like it should definitely help at the margin.

[sakulstra]

From my understanding the issue the aave protocol is facing is ETH withdrawals, not only borrowing. Either way it will probably reach 100% utilization at some point and the proposal does not much to stop it.

I’m not generally opposed to doing “sth”, but with the e.g. curve adjustments for me the ramifications are not clear. I feel like it might be best to do nothing

---

On a side note stopping eth borrowing might also be problematic for stETH positions, as there’s probably some dumping of stETH for ETH/powETH before the merge and by disabling borrowing it will be hard to keep up buy pressure on stETH (not my argument, but putting it here as didn’t see it mentioned before).

Disclaimer: As noted above i’m not experienced with risk and also didn’t do much on-chain analysis, plz take everything i write here with a big grain of doubt.

[Pauljlei]

@eboado, we appreciate the feedback - we hope the below is helpful:

On your first bullet point, Gauntlet received conflicting feedback from the community. Some stakeholders wanted to include the 60 AAVE payment, while others did not. This put us in a difficult spot because we 1) were driving the governance process but 2) we were not the actual party requesting the payment. In the end, we decided on the most democratic solution: in the Snapshot vote, we included all options (pay vs. not pay). Our AIP then reflects the decision of the community as represented by the Snapshot vote results. Should the community reach a clear consensus on retroactive funding (via a separate Snapshot vote or otherwise), we certainly agree that consensus should be followed. We also agree with you that retroactive funding is not a good precedent, but our AIP simply reflects the Snapshot vote’s decision.

We absolutely agree that speed is essential, especially with an event as important as the merge. But there are important considerations here:

• The implications of the merge are complex, especially for Aave, due to the unique stETH/ETH dynamics. As shown in this thread, there are many conflicting thoughts. Gauntlet has been analyzing the ETH merge for many weeks. It would be unacceptable to the community if Gauntlet rushes the analysis and then implements a proposal that adds risk to the protocol - the downsides of rushing are much higher than taking time to analyze properly. As stated above, there are risks with regard to changing the interest rate curve.
• The reason why we did not post in the public forums sooner than when Primoz did is very straightforward: there is front-running risk with this type of proposal. The proposal is to pause ETH borrow, and the governance process takes several days. If we gave advanced notice and broadcasted that Aave would be pausing ETH borrow, traders could possibly rush to borrow ETH ahead of the governance proposal, increasing ETH utilization and thus rendering the proposal useless. Of course, a potential consideration is that any sophisticated market participant would probably see it without it getting broadcast, and so we would only “surprise” the standard user, but nonetheless we shouldn’t amplify this risk.
• We were in contact with numerous stakeholders ahead of and during the original post. Also, we are unsure what it means by “it takes 7 days to answer them.” We answered just a few days afterward and posted a Snapshot vote immediately the day later.

As always, we appreciate your feedback. We are happy to discuss or answer any further questions.

[Emilio]

To me it’s a pretty simple matter: if borrowing doesn’t get disabled, there is pretty much certainty the weth pool will reach 100% utilization at some point. For borrowers, the risk of playing the ethpow game decreases as the merge approaches, becoming essentially 0 right before the merge. It is true that that this doesn’t in any way solve the withdrawal issue although it would be unwise to disallow withdrawals as it’s not even possible in v2 to granularly prevent this action from happening without freezing everything (and in general it would set a very bad precedent, users should always be free to withdraw their assets when there is liquidity available).
In general to me the potential risk of losing competitivity isnt relevant:

1. Borrowing should be reenabled immediately after the merge (and I hope gauntlet will plan the proposal to reactivate it accordingly)
2. As people withdraw, apy will take care of itself. Liquidity won’t increase (unlikely for that to happen anyway ahead of the merge) but apy should remain inline with historical and other competing protocols.

In general I am strongly in favor (I actually think this should have been done earlier) and I would go further and set a borrow cap for weth on all the v3 deployments as well l.

[Pauljlei]

Regarding the interest rate curve ramifications, Primoz’s calculations above are valuable but we’d add that the calculations and data provided underestimate the risk of stETH in several ways. In particular:

1. Users may swap stETH to ETH to farm pow ETH and swap it back after the merge, which will also affect the stETH peg.
2. The price slippage for stETH liquidations will increase if Curve LPs withdraw their stETH-ETH liquidity to farm pow ETH.

It is understandable that there is limited empirical data. These behaviors are hard to predict, and the impact of the behavior is difficult to estimate accurately. Although the tradeoffs of changing the interest rate curve are unclear, ETH supply should be monitored closely ahead of the merge.

[eboado]

Some extra analysis of mine

• If utilisation increases slightly more to 70% levels (highly possible), stETH/ETH leverage starts to become non-profitable. The increase of utilisation is so relatively small that, in practise, I don’t think there is any mechanism effective enough to stop it: if borrowing is halted, can easily be that some depositors will withdraw, so utilisation goes up. If the slopes of interest rate strategy are moved up, the “un-profitability point” of stETH/ETH is dragged down, so the pressure on stETH selling to ETH for unwinding, plus deposit rate of ETH goes down, and less incentive on ETH depositors.
  With so small margins on utilization regarding stETH/ETH, in my opinion, the scenarios depend on chaotic movements of borrowers and depositors: if stETH is strong enough, it should be fine, both on the stETH side and on stETH/ETH Aave side. Consequently, I would not even take this as a variable of the decision.

• In what affects withdrawals, I have a feeling that incentives/profitability based on basic Aave pool dynamics simply don’t matter in a situation like this. People understanding that looking for profit on some PoW fork is probably a good recipe to put their capital at risk should just not do anything with their Aave position, same as always. People looking for that kind of movement withdrawing ETH or closing more “complex” positions that lead to withdrawal of ETH, will just do it, no really matter the utilisation, rate, or whatever. Also considering that the 3% estimated value of ETH PoW is highly probably smoke, for sure in volume. A fork of Ethereum with PoW is a dead chain. In addition, stopping withdrawal is not and should never be a valid action, unless major technical risk. So regarding withdrawals, in my opinion, the same as with stETH/ETH: too chaotic, no effective mechanism → no action.

• Now, more sophisticated actors trying to borrow for relatively short amount of time. If the actor is really sophisticated, the borrow would happen just before the merge, and the closure of the borrowing just after, factually the risk for the protocol on that is close to 0; only a price update in the middle, for some reason dropping the ETH price maybe 20%. In my opinion, that carries 0 risks, so no action.
  For a bit less sophisticated actors borrowing more in advance and repaying later, they will need or 1) borrow with caution in terms of LTV → this will not increase so much utilisation 2) borrow with high LTV → price volatility of ETH highly probably will be higher → they will get burnt.
  So, the first group in my opinion carries no risk, and the second, if we assume a bit of rationality of the majority (not trying to risk capital to probably get nothing), is marginal → no action.

Also, we are removing importance on something critically important. On liquidations, even if utilization is 100%, it is possible to liquidate and take aWETH as liquidator, the same bonus as if taking underlying WETH. There is no reason to believe that liquidators will not think that taking aWETH is not profitable (I would personally do it).

So after following the discussion and giving a deep second thought, I will vote AGAINST stopping borrowing. Too chaotic assumptions, so usually is better to let the pool’s basic dynamics do their work.

[eboado]

@Pauljlei regarding the first point of the requested payment in-proposal, for me it seems a problem community-wise currently. Yes, I think simply Gauntlet shouldn’t have included even the option in Snapshot, and redirect the OP to the correct channel (Aave Grants DAO); but I don’t think Gauntlet did anything really wrong.
But the Aave community should start thinking a bit more about the direction of procedures, if not, decentralized coordination will become a mess down the line. Disclosure: I’m one of the community members that contacted Gauntlet to suggest at least a change on the initial Snapshot, including by default the payment.

Regarding the second point, the reality is that on 8th August, the BGD technical analysis was published here. 16th August the Aave Companies published a thread about the DAO taking a stance on the fork. August 23th, Block Analytica published its risk analysis. During this time, no news from Gauntlet. August 27th (correction: my mistake, not 7 days, but 3-4) Gauntlet answers Block Analitica.
I understand this is a complicated topic/decision, and precisely because of that I can’t really accept that the topic is not open to discussion way before, like 1 month at least, given that the risk aspects, in this case, are way more relative than any other.
Now we are in a situation where I’m pretty sure big stakeholders don’t even have time to form an educated opinion, and 1.5 days left of voting.

[stani]

Thanks for the OP for detailed overview of the topic and everyone who contributed to the discussion.

On “High ETH utilization potentially makes liquidations harder or impossible”:

Argument of harder to liquidate ETH borrow-based positions depends heavily on AMM/CEX liquidity, currently most AMM addresses are blacklisted so we might see some withdrawals of ETH from AMMs (might be worth analysing if it corresponds the increased ETH borrowing trends as well on Aave).

There are AMM pools where ETH liquidity will exist because of the nature of the pool such as the StkaBPT pool (Pool management - Balancer). However the above pool liquidity is at 100 mm USD (which of 20% in ETH) and we might we some withdrawals before The Merge (all though the 10 days cooldown buffer will help to quantify the exposure). Small example but demonstrates that there are pockets of ETH liquidity across AMMs and of course the liquidity in Cexes.

Would imagine that most of ETH used to repo collaterals would be recycled as long as there is ETH liquidity, would see less of an issue for the harder liquidations. Similarly as @eboado mentioned liquidators can also take aWETH, but I believe that most liquidators that operate in scale are not being long in their repo inventory - i.e. you sell whatever you liquidate and keep the proceeds. Smaller liquidators could be indeed interested in aWETH since they can hold it for a while period.

on “High ETH utilization increases the ETH rate to a level where stETH/ETH positions are making negative APY”

I definitely think reduction of exit liquidity for stETH might be a concern, however based on @Primoz analysis 10% stETH depeg does not yet cause devastating issues. Also want to highlight that the higher the temporary depeg, the wider arbitrage opportunity it creates (i.e. you have arb even on improvements on the peg, even if we won’t see a potential repeg until Shanghai).

On top of everything, the stETH liquidity is less relevant if there is no significant deleverages of the positions - in the current assessments even if the leveraged stETH positions become unprofitable for short-period time, that doesn’t necessarily result to unwind if the cost of borrowing returns to normal. Of course there is a cost for deleverage and you might actually want to wait (even during negative APY) for better market conditions for the actions.

In overall, what is important to consider is the timelines. For how long is the speculative behaviour going to happen and when does the unwinding back to the normal state occur time wise.

I believe that the conclusions of @MarcZeller and @sakulstra points also well out that stopping borrowing might actually push the liquidity away in overall elsewhere and create unfavourable circumstances.

Based on all the conversations and analysis I would vote for AGAINST stopping borrowing since I believe the fork speculation is a short time occurrence and there is no analysis that there ETH liquidity would dry up complete over that period of time.

Regarding the household items

• I don’t see an issue for the OP proposing “reward” for the proposal, in fact value should be rewarded and in this case the sentiment is snapshotted (the community can choose). I also don’t believe that Aave Grants DAO is the right place reimbursing for such work would rather see it paid directly from the DAO which leaves space for the AGD for budgeted grants and activities. Most importantly if we are able as a community to incentivise work around the community we would see more contributors and increased competition which leads to higher quality of engagements.
• I would liked have seen similar proposal from Gauntlet earlier (maybe even responding to OP on facts and leaving the debate going on while sharing own assessment). I don’t see the front-running risk as an issue as such front-run would anyways happen if happens during the voting period. I do also think that Gauntlet as major contributor to all things risk within the Aave community would favour in more lead in bringing the discussions to the community forum as well. However, I do not see any overlap since as we see the topic is quite hard and draws multiple conclusions and opinions. It’s very typical in risk assessment to have multiple opinions/assessments.
• Regarding setting precedents, we should actually be less cautious about proposals becoming as precedents to follow and leave room for the community to explore various options and alternatives regarding how to fund proposals and based on the outcomes understand more what works for the Aave community and bring the best results. For now would say we are quite in good position that we are able to attract contributors such as @Primoz and @monet-supply to help the Aave community and a small reward might motivate into more long-term, pro-founded work within the community similarly as Gauntlet and BGD is involved already.
• Community coordination is always something that is hard to get right but also given that our community has good foundations we should be in a good positions to set a good example for other communities and hence want to encourage every contributor to bring the best value from their engagements. We are all learning and improving, together.