Blocked Flops (Debt Auctions) - Everything You Need To Know

As was briefly discussed in Friday’s video call, there’s a small bug with the system’s accounting logic. Twenty fewer debt auctions, corresponding to 1 million DAI of debt, were triggered than expected on 3/19 and 3/22 as a result of this. Without action from governance, this unbacked debt cannot be properly auctioned and must instead await fees and/or liquidation penalties to nullify it.

Given the rather technical nature of the issue, this post will contain high-level information only; the full details can be found in a set of slides (to be presented at the next governance call).

Other important things to know:

  • further debt auctions are not blocked, just the 20
  • more auctions may become blocked (but at this point the only pending ones are the stuck ones)
  • Keeper logic can be tweaked to minimize the chances of this (auction-keeper repository already updated)
  • in order to fully recapitalize the system, governance should act promptly to unblock the remaining auctions
  • this is not necessarily an emergency (from a technical perspective), but it is important and remediation should not be unnecessarily delayed
  • in the long-term, smart contract updates can be made to ensure the issue does not reoccur (or can be fixed without the intervention of governance)–technical team already working on this

Open questions (not exhaustive):

  • How soon should governance unblock the auctions? (i.e. immediately, in next Friday’s executive, etc)
  • Should the spell to unblock also start the remaining auctions? (This ensures that the bug does not recur.)
  • Should the spell enforce a deterministic unblocking time (otherwise the auctions start whenever the spell is cast)?

Arguments for unblocking ASAP:

  • auction participants will be buying/holding DAI until auctions complete–with more than one bidder, this creates a multiplier effect that puts upward pressure on the peg
  • we don’t know how big the first effect is, but finishing the auctions removes one variable from our considerations of how to tune the rest of the system’s parameters
  • the sooner the auctions complete, the less risk there is to their success due to market volatility
  • avoids mixing technical fixes with monetary policy in the executive

This will probably be fixed earlier, but at the current rate Ash drops to zero in about 90 days.

I think both fixes are worth implementing:

  • drain Ash with suck and kiss
  • upgrade flop to kiss after first bid