CR141 — Governance Impact Dashboard

Replay of executed governance proposals against the block before execution. Observed events only.
cr141-v1.0-proof

Collateral factor changes

Route A / Route B cross-check

Route A derives the delta from the protocol's getHypotheticalAccountLiquidity response. Route B recomputes it independently from exchangeRateStored, getUnderlyingPrice and the CF arithmetic. Agreement within 1 wei means neither path drifted.
Disagreements are not value errors. Route A divides by cf_before, so integer truncation loses up to 1 wei per account; Route B does not truncate. Every disagreement in this corpus is that truncation artefact (CR141-INV-001), and the flip sets — the thing the verdict rests on — agree exactly: 474 predicted = 474 observed. A disagreement count here measures arithmetic rounding, not a modelling error.

All 28 proposals — click a column to sort

Density = predicted flips / targets. It is the discriminating metric: pid 295 flips 53 of 1,023 targets (5.18%), pid 71 flips 9 of 8,584 (0.10%). Ranking by raw flip count hides a 50× spread.
Every row is an account whose health status changes sign when the proposal executes. Predicted and observed are computed by two independent paths; a row where they disagree would render red. Corpus count today: 0.

All scenarios

Each row asks: what if this market's CF had been 5% higher or lower than what the proposal actually did? Affected counts accounts whose buffer moves. Newly flipped counts accounts that would break under the hypothetical but did not under the real change; healed is the reverse. Only the named market changes — every other market keeps its actual cf_after.
These are computed, not estimated. The simulator reuses the same arithmetic the audit used, and its self-check requires that asking for the actual CF reproduces the audit's predicted to the wei. Borrow impact is 0 by construction: a CF change moves collateral, not debt.

Audit gates

Exactness

Reproduce offline

cd rerun_package && python3 verify_hashes.py