Thanks for putting these out for review. We spent most of our time in lending.md, and section 7 (Open Design Questions) is the part closest to our own work. On Compound we implemented the CAPO oracle for the correlated-collateral case, built the reserve-growth tracking, and we operate Compound Labs’ Configurator pattern on the markets we deploy. Several of the section 7 questions are things we handle in production.
One observation from that side: oracle-induced risk and market risk fail differently and want separate controls. Your maxStaleness field already points that way. We wrote up the correlated-collateral case here: Correlated collateral and the parameter layer: notes for Canton vault builders
The question we’d like your read on is scope. Buffer sizing, debt ceilings, the change path for VaultParams, stress evidence for the insurance fund: is any of that planned for the reference implementations in M2 or M3, or does it stay a layer that implementers bring? We ask because that layer, parameter governance and calibration on Zenith, is what our open Dev Fund proposal covers (Proposal: Canton Risk Engine by Noosphere-314 · Pull Request #665 · canton-foundation/canton-dev-fund · GitHub). If you’re already building it we’d rather know now. If you aren’t, we’d rather compose with your stack than build something that sits awkwardly next to it.
Glad to go further on any of this, in whatever format is useful to you.