Skip to main content
Pharos

Redemption Backstop Changelog: v2 archive

Historical release evidence. Return to current versions and the complete history index.

What this controls

Redemption Backstop Changelog: v2 archive version history

Use this changelog to trace how Pharos changed the related scoring, data-source, threshold, or interpretation rules over time. Entries record the contract active at each release; archived rules are retained so historical charts and citations can be read against the rules active at the time.

Latest version

v2.9Mar 24, 2026

Semantics correction for non-deterministic HOLLAR exit

A route-semantics review removes one overstated redemption path and explicitly leaves several harder assets outside medium-confidence coverage until a credible holder backstop is established.

Impact snapshot

  • HOLLAR is no longer modeled as a `psm-swap` redemption route because the Hydration Stability Module only guarantees buying HOLLAR from the facility, while protocol buybacks of HOLLAR remain opportunistic rather than holder-deterministic
  • The harder follow-up set led to no new medium-confidence additions for crvUSD, sUSD, MIM, or USDU Finance because current public materials still do not establish a primary redemption rail comparable to the existing modeled route families
  • This keeps redemption coverage honest by preferring uncovered or low-coverage states over overstated direct-exit semantics
v2.9Mar 24, 2026

Semantics correction for non-deterministic HOLLAR exit

A route-semantics review removes one overstated redemption path and explicitly leaves several harder assets outside medium-confidence coverage until a credible holder backstop is established.

  • HOLLAR is no longer modeled as a `psm-swap` redemption route because the Hydration Stability Module only guarantees buying HOLLAR from the facility, while protocol buybacks of HOLLAR remain opportunistic rather than holder-deterministic
  • The harder follow-up set led to no new medium-confidence additions for crvUSD, sUSD, MIM, or USDU Finance because current public materials still do not establish a primary redemption rail comparable to the existing modeled route families
  • This keeps redemption coverage honest by preferring uncovered or low-coverage states over overstated direct-exit semantics
Details

Impact notes

  • HOLLAR is no longer modeled as a `psm-swap` redemption route because the Hydration Stability Module only guarantees buying HOLLAR from the facility, while protocol buybacks of HOLLAR remain opportunistic rather than holder-deterministic
  • The harder follow-up set led to no new medium-confidence additions for crvUSD, sUSD, MIM, or USDU Finance because current public materials still do not establish a primary redemption rail comparable to the existing modeled route families
  • This keeps redemption coverage honest by preferring uncovered or low-coverage states over overstated direct-exit semantics
Commit provenance not recorded
    v2.8Mar 24, 2026

    Second medium-confidence redemption cleanup tranche

    A second cleanup tranche upgrades the best non-top-100 low-confidence routes where Pharos already had sufficient issuer, reserve, or queue-redemption evidence to stop relying on heuristics.

    • cUSD, cEUR, ALUSD, and AZND now use reviewed eventual or reserve-backed redemption semantics instead of heuristic capacity ratios
    • USDA now carries a reviewed issuer-redemption route, while pUSD Plume now uses live reserve metadata with a documented 1:1 USDC fallback instead of a generic low-confidence issuer assumption
    • Names whose route semantics are still genuinely unresolved, such as crvUSD, sUSD, MIM, and HOLLAR, remain outside this tranche rather than being promoted on weak evidence
    Details

    Impact notes

    • cUSD, cEUR, ALUSD, and AZND now use reviewed eventual or reserve-backed redemption semantics instead of heuristic capacity ratios
    • USDA now carries a reviewed issuer-redemption route, while pUSD Plume now uses live reserve metadata with a documented 1:1 USDC fallback instead of a generic low-confidence issuer assumption
    • Names whose route semantics are still genuinely unresolved, such as crvUSD, sUSD, MIM, and HOLLAR, remain outside this tranche rather than being promoted on weak evidence
    Commit provenance not recorded
      v2.7Mar 24, 2026

      Buffer-backed medium-confidence redemption tranche

      A follow-up tranche promotes the remaining cleanest heuristic routes by tying their capacity bounds to already-curated stable redemption buffers or direct full-reserve rails.

      • USDD, LISUSD, reUSD, USR, USDF, DUSD, USP, and BUCK now use reviewed documented-bound capacity instead of generic heuristic ratios because Pharos already tracks explicit stable redemption buffers for those routes
      • msUSD and fxUSD now carry reviewed direct-redemption semantics rather than unresolved low-confidence defaults, reflecting Main Street's full USDC reserve rail and f(x)'s documented collateral redemption path
      • Routes whose reserve stack still lacks a clearly bounded redeemable stable buffer, such as YUSD, USN, and UTY, intentionally remain low-confidence until the evidence improves
      Details

      Impact notes

      • USDD, LISUSD, reUSD, USR, USDF, DUSD, USP, and BUCK now use reviewed documented-bound capacity instead of generic heuristic ratios because Pharos already tracks explicit stable redemption buffers for those routes
      • msUSD and fxUSD now carry reviewed direct-redemption semantics rather than unresolved low-confidence defaults, reflecting Main Street's full USDC reserve rail and f(x)'s documented collateral redemption path
      • Routes whose reserve stack still lacks a clearly bounded redeemable stable buffer, such as YUSD, USN, and UTY, intentionally remain low-confidence until the evidence improves
      Commit provenance not recorded
        v2.6Mar 23, 2026

        Moderate-effort redemption confidence tranche

        A moderate-effort tranche reviews a final group of already-modeled lower-confidence routes where Pharos now has stronger primary redemption semantics, but not yet protocol-native live instant-buffer telemetry across the full set.

        • DOLA and JupUSD now treat their published stable-buffer bounds as reviewed documented-capacity inputs instead of leaving those ratios in the heuristic bucket
        • rwaUSDi, mTBILL, MUSD, USDN, and YZUSD now use reviewed redemption semantics with documented-bound capacity instead of generic low-confidence placeholders, while YUSD, USN, and UTY keep conservative bounded-capacity assumptions because their delta-neutral collateral stacks still lack explicit published live buffers
        • The documented-bound subset now contributes medium-confidence redemption evidence, while the reviewed delta-neutral routes stay visible-only until Pharos has explicit buffer bounds or live telemetry
        Details

        Impact notes

        • DOLA and JupUSD now treat their published stable-buffer bounds as reviewed documented-capacity inputs instead of leaving those ratios in the heuristic bucket
        • rwaUSDi, mTBILL, MUSD, USDN, and YZUSD now use reviewed redemption semantics with documented-bound capacity instead of generic low-confidence placeholders, while YUSD, USN, and UTY keep conservative bounded-capacity assumptions because their delta-neutral collateral stacks still lack explicit published live buffers
        • The documented-bound subset now contributes medium-confidence redemption evidence, while the reviewed delta-neutral routes stay visible-only until Pharos has explicit buffer bounds or live telemetry
        Commit provenance not recorded
          v2.5Mar 23, 2026

          Reviewed docs-backed quick-win redemption tranche

          A docs-backed quick-win tranche upgrades nine existing low-confidence redemption routes where the remaining blocker was heuristic capacity or stale access and fee assumptions rather than missing telemetry.

          • avUSD, cUSD, USDu, cgUSD, HONEY, EUSD, AID, OUSD, and USBD now use reviewed documented-bound redemption capacity instead of staying low-confidence under heuristic supply models
          • USDu and AID now reflect whitelist-gated direct redemption access, while cgUSD and AID also disclose reviewed live fee assumptions from official docs
          • These routes still do not claim a separately measured live instant buffer, but they now contribute medium-confidence redemption evidence across roughly half a billion dollars of additional tracked market cap
          Details

          Impact notes

          • avUSD, cUSD, USDu, cgUSD, HONEY, EUSD, AID, OUSD, and USBD now use reviewed documented-bound redemption capacity instead of staying low-confidence under heuristic supply models
          • USDu and AID now reflect whitelist-gated direct redemption access, while cgUSD and AID also disclose reviewed live fee assumptions from official docs
          • These routes still do not claim a separately measured live instant buffer, but they now contribute medium-confidence redemption evidence across roughly half a billion dollars of additional tracked market cap
          Commit provenance not recorded
            v2.4Mar 23, 2026

            Maple syrup withdrawal route correction

            Maple's syrupUSDC and syrupUSDT routes now model the documented withdrawal queue instead of an overstated near-instant redemption buffer.

            • syrupUSDC and syrupUSDT now use reviewed queue-redemption semantics with documented-bound eventual capacity rather than a heuristic 30% immediate buffer assumption
            • Access is now modeled as whitelisted onchain, reflecting Maple's PoolPermissionManager gating for `requestRedeem` and `redeem` calls
            • These routes now contribute medium-confidence redemption evidence while preserving Maple's documented FIFO processing and potential multi-day settlement delay
            Details

            Impact notes

            • syrupUSDC and syrupUSDT now use reviewed queue-redemption semantics with documented-bound eventual capacity rather than a heuristic 30% immediate buffer assumption
            • Access is now modeled as whitelisted onchain, reflecting Maple's PoolPermissionManager gating for `requestRedeem` and `redeem` calls
            • These routes now contribute medium-confidence redemption evidence while preserving Maple's documented FIFO processing and potential multi-day settlement delay
            Commit provenance not recorded
              v2.3Mar 23, 2026

              Reviewed lower-cap redemption cleanup tranche

              A small lower-cap cleanup tranche upgrades Pleasing and Apyx routes from generic heuristics to reviewed redemption semantics without adding new live telemetry assumptions.

              • PUSD and PGOLD now use reviewed documented-bound redemption routes tied to Pleasing's published off-ramp and physical-delivery docs instead of generic heuristic issuer assumptions
              • apxUSD now reflects the documented whitelist-gated mint/redeem rail rather than a generic permissionless stablecoin-redeem assumption
              • These routes still do not claim a separately measured live instant buffer, but they now contribute medium-confidence redemption evidence instead of remaining low-confidence heuristics
              Details

              Impact notes

              • PUSD and PGOLD now use reviewed documented-bound redemption routes tied to Pleasing's published off-ramp and physical-delivery docs instead of generic heuristic issuer assumptions
              • apxUSD now reflects the documented whitelist-gated mint/redeem rail rather than a generic permissionless stablecoin-redeem assumption
              • These routes still do not claim a separately measured live instant buffer, but they now contribute medium-confidence redemption evidence instead of remaining low-confidence heuristics
              Commit provenance not recorded
                v2.2Mar 23, 2026

                Live-buffer routes for Ethena and Falcon synthetics

                USDe and USDf now reuse live reserve telemetry for current redeemable stable buffers, turning two large synthetic-dollar gaps into reviewed route coverage.

                • USDe now models the whitelisted direct mint-and-redeem rail documented by Ethena, with fresh live Liquid Cash telemetry used as the current redeemable stable buffer and a conservative 0.5% fallback bound when telemetry is unavailable
                • USDf now models Falcon's KYC-only queued redemption route with a live stablecoin-buffer input from Falcon's transparency feed and a reviewed zero protocol-fee assumption based on Falcon docs
                • These routes materially expand medium-confidence redemption coverage without pretending either protocol has a permanently fixed instant-exit buffer
                Details

                Impact notes

                • USDe now models the whitelisted direct mint-and-redeem rail documented by Ethena, with fresh live Liquid Cash telemetry used as the current redeemable stable buffer and a conservative 0.5% fallback bound when telemetry is unavailable
                • USDf now models Falcon's KYC-only queued redemption route with a live stablecoin-buffer input from Falcon's transparency feed and a reviewed zero protocol-fee assumption based on Falcon docs
                • These routes materially expand medium-confidence redemption coverage without pretending either protocol has a permanently fixed instant-exit buffer
                Commit provenance not recorded
                  v2.1Mar 23, 2026

                  Mid-cap route correction and review tranche

                  A mid-cap tranche adds missing USX, USDa, and M redemption configs while correcting USD.AI and NUSD onto reviewed routes that better match their protocol docs.

                  • USX, USDa, and M now carry reviewed redemption routes instead of remaining uncovered, and NUSD now uses reviewed documented-bound queue semantics rather than a generic 20% heuristic
                  • USD.AI now models the base token's direct burn-and-withdraw stablecoin rail instead of inheriting the slower sUSDai unstaking assumptions
                  • These routes still do not claim a separately measured live instant buffer, but they materially expand medium-confidence redemption coverage across the mid-cap queue
                  Details

                  Impact notes

                  • USX, USDa, and M now carry reviewed redemption routes instead of remaining uncovered, and NUSD now uses reviewed documented-bound queue semantics rather than a generic 20% heuristic
                  • USD.AI now models the base token's direct burn-and-withdraw stablecoin rail instead of inheriting the slower sUSDai unstaking assumptions
                  • These routes still do not claim a separately measured live instant buffer, but they materially expand medium-confidence redemption coverage across the mid-cap queue
                  Commit provenance not recorded
                    v2.0Mar 23, 2026

                    Third lower-cap redemption review tranche

                    A third lower-cap review tranche upgrades more issuer-style routes from heuristic supply-full modeling to reviewed documented-bound capacity and corrects frxUSD onto its direct onchain stablecoin redemption rail.

                    • thBILL, XAUm, USDGO, and USA₮ now carry reviewed documented-bound eventual redemption capacity instead of generic heuristic supply-full modeling
                    • XAUm now discloses a reviewed 25 bps redemption fee and T+3 settlement expectations, while USDGO now uses the reviewed zero-fee StableHub exchange rail documented by OSL
                    • frxUSD now models the direct onchain USDC mint/redeem contract path as a reviewed stablecoin-redeem route instead of sitting in a generic offchain issuer bucket
                    Details

                    Impact notes

                    • thBILL, XAUm, USDGO, and USA₮ now carry reviewed documented-bound eventual redemption capacity instead of generic heuristic supply-full modeling
                    • XAUm now discloses a reviewed 25 bps redemption fee and T+3 settlement expectations, while USDGO now uses the reviewed zero-fee StableHub exchange rail documented by OSL
                    • frxUSD now models the direct onchain USDC mint/redeem contract path as a reviewed stablecoin-redeem route instead of sitting in a generic offchain issuer bucket
                    Commit provenance not recorded