Limitations¶

One section per row of the support matrix's limitations: what this version does not do, or does only under conditions. A not-supported explanation names its limitation. How to read an explanation, and how to ask for one (openflowsheet explain <code>, the explain operation), is on Messages.

`L-COL-1`: L-COL-1¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: The column's feeds are single-phase and declared: one side feed per stage, none at a slot stage (S01 DECISION 8, S02 §1.4), and every side feed on its feed port shares one declared phase (the revision reader's one phase per port). A two-phase feed needs a flash and two stages.

`L-COL-10`: L-COL-10¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: A column under a PTC policy (T08-ptc-v1) ends PTC_MAPPING_INVALID before any attempt: the PTC mapping does not cover the column's stage variables.

`L-COL-2`: L-COL-2¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: The column has no side draws, pump-arounds, stage efficiencies or hydraulics, and is solved by S02's Newton method only: no inside-out or rate-based method (S02 §14).

`L-COL-3`: L-COL-3¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: A total condenser gives saturated liquid (no subcooling); there is no total reboiler.

`L-COL-4`: L-COL-4¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: On wilson-ig-v2 the column's end specifications are ratio, rate and duty only: a purity, recovery or temperature end is refused at binding (S10-C D8).

`L-COL-5`: L-COL-5¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: No uniqueness, multiplicity or feasibility claim for a column beyond S06 §14: a certificate holds at the converged state, and an inverse specification's feasibility is decided only on S06's registered band.

`L-COL-6`: L-COL-6¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: In a flowsheet a purity end's tolerance acts on the mole fraction directly (|x − x*| ≤ the molar-flow tolerance ÷ 1 mol/s, the F_ref meaning); the standalone column's acts on F·x.

`L-COL-7`: L-COL-7¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: On wilson-ig-v2 the column's Stage B (S08's block re-entry) is tested at the opening level only (S10-C).

`L-COL-8`: L-COL-8¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: SYN-001 columns are synthetic (L18): their results are verified, never validated.

`L-COL-9`: L-COL-9¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: In a flowsheet a column region has edge-3 continuation only under RRC-revision-v1, from a verified ancestor that differs only in operating values. A first solve, a structural edit (stage count, feed stage, a changed end-specification choice) and the default policy have none. The unit's own evaluate keeps S02's full standalone path, S08 Stage B included, so the traversal start is the standalone answer.

`L-CSTR-1`: L-CSTR-1¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: The kinetic CSTR's rate law and coolant model are synthetic (the Frank-Kamenetskii γ → ∞ form and a coolant capacity rate); no Arrhenius or physical-conductance claim is made.

`L-CSTR-2`: L-CSTR-2¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: An adiabatic, dormant kinetic CSTR's outlet temperature is a singularity (T05 §4.7 (a)); v0.1 does not certify an EO solve in which it appears.

`L-CSTR-3`: L-CSTR-3¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: The kinetic CSTR's causal evaluator is a local model initializer (the isothermal state at T_init), not a steady state.

`L-E01-1`: L-E01-1¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Since R-898 every P2 capital row is floored, so P2's optimization exercises operating cost only; no registered NLP case carries a capital row in a units:N regime. A floor's derivative row is empty (E01.A18), so E01.A33–A36 show nothing about capital sensitivities through the optimizer. P2 is not resized (E01 design §9.7).

`L-GP-1`: L-GP-1¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: On Wilson (wilson-ig-v2), only the five K02/T05 units and the two machines; no pump (no liquid volume); exchanger, separator and reactors refused.

`L-GP-2`: L-GP-2¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Wilson machines are single-phase vapour, within the records' domain (300–400 K, 30–200 kPa).

`L-GP-3`: L-GP-3¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: A Wilson TP flash near an azeotrope has an E-2 noise floor (S10 §14); a temperature-degenerate split is judged on its stored split.

`L-OPT-1`: L-OPT-1¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Revision optimisation (ADR 0071) runs only on the revision_eo route in the audited nlp environment on linux-x86_64. Each start holds its phase and assessment regimes: a candidate whose regime changed is NOT_VERIFIED; a boundary optimum is verified only on the single-phase side of a guarded regime (phase_boundary_form, flagged, and a plain solve of the committed candidate certifies UNVERIFIED); two-phase-side and unguarded-regime boundary optima and candidates in the rank-limited-but-inactive gap are NOT_VERIFIED; the zero-flow subsystem is held or unresolved per start, all or nothing. KKT_POINT_VERIFIED is numerically verified local stationarity, not optimality.

`L-OPT-2`: L-OPT-2¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Ipopt runs with fixed options (monotone barrier, limited-memory Hessian). On a nearly linear objective whose constraint must become active from inside the feasible region it can stall at max_iter (measured on OPT-L from 673.15 K and 661 K; it verifies from 650 K, a start that violates the constraint). Such a start is reported SOLVER_FAILED, never verified; another start is the remedy.

`L-S11-1`: L-S11-1¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: iapws95-v1 (IAPWS-95, S05) is selected by record_source: benchmarks/s05/components.yaml for the one component H2O, over 273.16–1273.15 K and 10²–10⁸ Pa. The routing models (syn001.feed_source, syn001.product_sink, syn001.stream_splitter) and the seven thermal.* models (L-TH-1..8) bind on it; its evidence is S05's validation against the IAPWS-95 reference equation (S05 spec).

`L-S11-2`: L-S11-2¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: coolprop-heos-v1 (CoolProp HEOS, S07) is selected by record_source: benchmarks/s07/components.yaml for exactly one admitted fluid per revision, among the 123 of S07's table; each fluid keeps S07's tier and domain. The routing models and the seven thermal.* models (L-TH-1..8) bind on it, and it needs the coolprop extra: without CoolProp installed, binding is refused basis_unavailable(coolprop-heos-v1).

`L-S11-3`: L-S11-3¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: wilson-ig-v2 (S10's Wilson activity model with an ideal-gas vapour; wilson-ig-v1's answers with the molar entropy added, S11 units design §4.6) is selected by record_source: benchmarks/s10/records.yaml for its three record components (CH3OH, C2H5OH, H2O, in any order), over 300–400 K and 30–200 kPa. The routing models, the column (syn001.column, L-COL-1..10), the five K02/T05 units with γ–φ faces and the vapour machines thermal.compressor and thermal.turbine (L-GP-1..3) bind on it. Its records ship as package data (R-880), so it binds in a wheel. An installation with neither copy is refused basis_unavailable(wilson-ig-v2). The library constructors WilsonIdealGasProvider() and WilsonIdealGasProviderV2() with no argument read a source checkout only and elsewhere refuse records_missing. A library user passes resources.packaged_path(wilson_ig.RECORDS_PATH) (R-883). Premises P1–P3 are checked at binding. The S10 VAL-1–3 bands apply, and the VAL-3 latent-heat miss (methanol, +5.72 % against 5 %) is carried. Results are compared across architectures by tolerance, not bitwise.

`L-TH-1`: L-TH-1¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Pure-fluid stream regimes are declared (liquid, vapor, vapor_liquid = saturated); a closure in another regime is refused phase_regime_mismatch.

`L-TH-2`: L-TH-2¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Saturated pure-fluid streams are produced only by thermal.* units; there are no wet feeds, and routing models do not carry them.

`L-TH-3`: L-TH-3¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: thermal.flash needs a two-phase state at drum pressure; its products are declared vapor_liquid.

`L-TH-4`: L-TH-4¶

Kind: Notice: not a failure.

Summary: A turbine's ideal (isentropic) outlet is computed in the phase declared for its real outlet, so a vapour outlet whose ideal state would be wet is refused.
Why: This version takes the phase of a machine's ideal outlet from the phase declared for its real outlet. It does not compute a two-phase ideal state for a turbine whose outlet is declared vapour, and it refuses a compressor whose ideal state would be two-phase.
Do: Declare the outlet wet if it can be wet at your conditions, or change the operating conditions until the ideal and the real outlet are in the same phase; openflowsheet explain isentropic_state_two_phase gives the details.
See also: isentropic_state_two_phase

Example:

A turbine's ideal (isentropic) outlet is computed in the phase declared for its real outlet, so a vapour outlet whose ideal state would be wet is refused.
  Why:     This version takes the phase of a machine's ideal outlet from the phase declared for its real outlet. It does not compute a two-phase ideal state for a turbine whose outlet is declared vapour, and it refuses a compressor whose ideal state would be two-phase.
  Do:      Declare the outlet wet if it can be wet at your conditions, or change the operating conditions until the ideal and the real outlet are in the same phase; `openflowsheet explain isentropic_state_two_phase` gives the details.
  Kind:    Notice: not a failure.
  Code:    L-TH-4 (openflowsheet explain L-TH-4)
  Docs:    https://computational-chemical-engineering.github.io/openflowsheet/messages/limitations/#L-TH-4

`L-TH-5`: L-TH-5¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: The near-critical box and the supercritical label rule of S05 §4.2 and S07. A supercritical PH or PS closure whose root lies at the provider's label switch (T = T_c) returns not_converged (measured: 4 of 11 200 grid outcomes, at 22.85 MPa).

`L-TH-6`: L-TH-6¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: No regime discovery for pure fluids.

`L-TH-7`: L-TH-7¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Dormant single-phase thermal outlets need T05b's v2 phase contract to solve. Under v1 a converged start certifies RANK_DEFICIENT, and a solve that must move ends LINEAR_SOLVE_FAILED. Under either contract, a converged run with such an outlet certifies UNVERIFIED, with regularity RANK_DEFICIENT. The verifier's own dormant-outlet table (verify.zero_flow.DORMANT_OUTLETS, spec §7.6) has no thermal.* outlet, so it judges the outlet's energy row, which reads no temperature at zero flow.

`L-TH-8`: L-TH-8¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: A revision declares one phase per port (R4), so a thermal.mixer takes inlets of one declared phase. Mixing a liquid with a vapour is available by direct construction only, as L-COL-1 is for the column's side feeds.

`L-WEB-1`: L-WEB-1¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: The web shell (serve-http --ui) is a diagnostic client of the HTTP binding: it edits nothing (Solve and Cancel are its only effects), calls eighteen operations and no other, computes only display arithmetic (design note §5.6), converts no unit, and judges nothing — the scenario view displays run differences without a registered comparison or allowance.

`L-WEB-2`: L-WEB-2¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: The web shell's security is the local, single-user model: loopback by default, the bearer token on every data request, a strict Content-Security-Policy; no TLS, no multi-user hosting, and no security assessment beyond design note §7. Served with --allow-remote over plain HTTP the page is an insecure context and record digests show "unchecked".

`L-WEB-3`: L-WEB-3¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: The web shell is tested in headless Chromium only. No usability, workflow or screen-reader study exists; accessibility is limited to status shown as glyph and word, controls that are keyboard-reachable by construction, and the measured contrast of its colour tokens.

`L-WEB-4`: L-WEB-4¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: The agent history shows what the project recorded, not what an agent claimed; the validation shown beside a commit is recomputed now by the current server, not the report the agent received at commit time.

`L-WS-1`: L-WS-1¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Warm-start compatibility is exact id-set identity; a topology edit is a typed rejection to the next source, not a partial or adapted match.

`L-WS-2`: L-WS-2¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: The warm-start source is chosen automatically (the lineage's latest VERIFIED state) under an opt-in policy only; no request-named source exists in v0.1 (v0.2, M03).

`L-X04-1`: L-X04-1¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Gauge pressures are not accepted as inputs: ADR 0016's input rows stand, and a gauge unit (barg, psig) in a revision or a change is refused with its existing code. Gauge units are display units only, resolved against the project's declared ambient pressure. Input conversion at commit is the planned remedy, not built in 0.4.

`L-X04-2`: L-X04-2¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Display units change responses only. Stored and hashed documents, raw artifacts, check quotes and the web shell's as-solved state table stay SI; a display preference never reaches a run, a bundle, a certificate or a cache key.

`L-X04-3`: L-X04-3¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: A number without a display kind or a display row stays SI under every preference: the stream table's S_flow (W/K), V_flow (m³/s), rho (kg/m³), v (m³/mol) and molar_mass, and every message number without a quantity kind.

`L01`: L01¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: PTC is experimental: selectable by explicit policy only (T08-ptc-v1); no qualified family. V14 clause (b) is carried as FAIL (ADR 0021 D3; on PTC-R1, PTC ends at the saddle from 149 starts).

`L02`: L02¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: A zero-duty lifted flash whose root lies exactly on its dew or bubble point is UNVERIFIED, RANK_DEFICIENT; nearby two-phase-attempt stops are caught FAILED(false_success_detected).

`L03`: L03¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: The alias-pressure-shift limit: the pressure column at 0-based position j is certifiable iff 997 (j + 1) <= max(200 000 - P, P - 50 000) Pa; beyond it the alias certificates are unsupported(pressure_shift_outside_domain) and the verdict UNVERIFIED (spec §5.4).

`L04`: L04¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: A saturation-closure failure inside the rows' own window is FAILED, where UNVERIFIED(label_not_demonstrated) would be exact (conservative).

`L05`: L05¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: VERIFIED certifies residuals at registered tolerances; the solution-error bound is recorded, not promised.

`L06`: L06¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Energy checks share the property provider with the solver ([A09]).

`L07`: L07¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Bitwise float reproduction is reported, never promised; per-start ensemble outcomes differ across architectures. LU pivot-path diagnostics (u_diag_min_abs, u_diag_max_abs, u_diagonal_ratio, nnz_L, nnz_U) are recorded and not compared across platforms (ADR 0025 D4).

`L08`: L08¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Property-call counts are within-platform diagnostics; the terminal-refinement core (ADR 0018) can exhaust the property budget where newton finishes.

`L09`: L09¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: The robustness claim is the nominal profile only; the stress profiles were not run.

`L10`: L10¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: The initializer chain is partial (blueprint §7.4, K03 §10.1): GUESS-role specifications are not bound by the revision binder (such revisions run on legacy_eo), the SYN-001 tear path has its registered initializer only, and the one fallback between initializer sources is T08-warm-v1's compatible warm start → traversal (opt-in; L-WS-1, L-WS-2).

`L11`: L11¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Cancellation: verify has no cooperative checkpoint; a region's events are lost on interrupt.

`L12`: L12¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Agent evidence: one pinned model and configuration, MCP only, not out of sample (CP lower bound 0.681); T09 0/6, T10 1/3; INJ-4 never exposed.

`L13`: L13¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: worker.log is readable through get_artifact.

`L14`: L14¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: The v0.1 rank rule is exactly one freed column per promoted target.

`L15`: L15¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Linux x86-64 and aarch64 only; Python 3.13; the one lock. The C1 reactor (c1.reactor) is measured on x86-64 only (L44).

`L16`: L16¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: IDAES agreement holds at the registered near-sharp SmoothVLE ε only.

`L17`: L17¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: The reference environment's DWSIM runs on .NET 8, end of life 2026-11-10.

`L18`: L18¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Nothing in v0.1 is empirically validated: SYN-001 is synthetic (pseudo-components, invented constants, ideal VLE), and every result is numerical verification against the same equations.

`L19`: L19¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Certificates are local: the regularity screen and the error bound hold at the certified state only. No root completeness, uniqueness, thermodynamic stability beyond the admissibility check, or dynamic stability is assessed (claims stay NOT_ASSESSED).

`L20`: L20¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Convergence is not guaranteed wherever a root exists: registered counterexamples end typed (PHS-05 without edge 3; HOM-03/HOM-04 stall at a phase boundary); folds are not followed (no arclength continuation).

`L21`: L21¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Edge 3's continuation exists for a region with a promoted specification (specification continuation). Under RRC-revision-v1, it also exists for a revision region whose nearest verified ancestor has the same declaration and differs only in registered operating values (revision continuation, ADR 0072). Everywhere else edge 3 is the sequential restart or unsupported(…). Neither continuation changes the active set. RRC-revision-v1 is not a superset of T06-revision-v2: a continuation that stops part-way ends the step and the restart is not tried (one recovery per step), so it can fail where T06-revision-v2 converges, and no registered case is rescued by it that T06-revision-v2 does not rescue.

`L22`: L22¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Scope of the solver evidence: tear selection is greedy (minimality not claimed); behaviour finer than the registered basin margins and trajectories below a residual of 1e-6 are not established; several stated rules are exercised only by synthetic fixtures.

`L23`: L23¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: More than one recycle loop or region per plan is not exercised by any registered flowsheet; the plan builder's single-loop reading is to be revisited when one is (T02 review N8).

`L24`: L24¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Not implemented: Wegstein and Aitken acceleration, trust regions, Krylov linear solves, arclength continuation.

`L25`: L25¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Alias certificates cover pressure and temperature copies only (the kinds ADR 0001 D6 gives a tolerance); any other affine redundancy is uncertified_affine, retained and listed.

`L26`: L26¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: SYN-001's symmetric data leave paths unexercised: component discrimination in liquid enthalpy, a liquid heat of reaction (every liquid reaction is thermoneutral), an exchanger with a phase change or temperature-dependent c_p (refused).

`L27`: L27¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: An EO solve with a dormant PH-type outlet is never certified (UNVERIFIED, RANK_DEFICIENT; T05 §4.7 (a), R-050).

`L28`: L28¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: A single flowing component inside its latent jump is solved causally only; on the EO path it ends typed or not VERIFIED, never VERIFIED (T05 §4.7 (b), R-050).

`L29`: L29¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: The PH kernel's accuracy for near-pure feeds is the energy tolerance's resolution; a target inside the latent jump may be refused ph_ill_conditioned (T05 §4.4, R-049).

`L30`: L30¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Degenerate specifications are reported, not certified: a TP outlet of one component at exactly T_sat, conditioning near β → 0 or 1, and a zero-flow SPECIFICATION_CONFLICT (not searched past); a dormant exchanger side gets no second-law judgement.

`L31`: L31¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: A state whose verifier projection is refused is judged at its final state: a possible false alarm, never a false pass.

`L32`: L32¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: The ensemble gate's run 2 is not out of sample for T06-revision-v2 (its last change was informed by these draws); the holdout (436 of 440 on every class) is the out-of-sample estimate.

`L33`: L33¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Scope defaults of T06 §16: results are keyed by component id; a TWO_PHASE split with a vanishing phase is not converted by the solver (the verifier catches the label); three F-BUDGET starts have no cost remedy.

`L34`: L34¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: A replay bundle references the CasADi wheel by hash and does not embed it; a replay needs the wheel to remain fetchable.

`L35`: L35¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: A tear-path solve that exhausts its budget records attempts = 0, the start state and no provenance (K03's registered capped-budget shape).

`L36`: L36¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: No performance claim at size: costs are within-platform measurements on registered cases of at most tens of variables; compiles per plan run and the regularity screen's cost at size are not measured.

`L37`: L37¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: verify is not safe to run concurrently in threads of one process: the regularity screen seeds and restores numpy's global generator around onenormest (T06 §8.1, R-069), so concurrent calls can move the estimate, rcond₁, the bound and a regularity verdict near its threshold; jobs run in separate processes and are unaffected.

`L38`: L38¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: Structural analysis is generic: validation's STR checks and inspect_structure report structural rank and the DM/BTF partition only; numerical rank, conditioning and feasibility are judged by the certificate's regularity screen at a converged state (blueprint D05).

`L39`: L39¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: The structural-zero release (recovery edge E5) leaves no trace record: a trace does not show that it ran.

`L40`: L40¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: The reference-tool data (the IDAES and DWSIM environments, their downloads and their outputs) and OpenIDAES-450 are not distributed: the sdist and the wheel carry the project's own code, its package data (since M01 including the five C1 records, M01 spec §15 Q-N4's default), its licence files, one third-party data file, the IAPWS-95 coefficients benchmarks/s05/iapws95_coefficients.json, under IAPWS's terms with their attribution in NOTICE (S05 spec §16, R-616, R-617), and since E01 six third-party assessment data files under assessment/ (the SSLW cost set transcribed from IDAES under its BSD licence, the PrISMa chain factors, the BLS, Eurostat and Federal Reserve index series, and five IPCC AR6 GWP100 values), each with its attribution in NOTICE (ADR 0052), and since 0.3.0 Wilson's records benchmarks/s10/records.yaml and the Kurihara 1993 Table II points benchmarks/s10/vle/kurihara1993_binaries.csv, with their attribution in NOTICE (V03 spec §9.1, R-880), and nothing else; every ComponentRecord is synthetic (SYN-001) or one of M01's five real C1 records, each with its identifiers, rights, per-parameter provenance and retrieval equality (M01 spec §3.5, R-158).

`L41`: L41¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: A bundle written by an installed package or a checkout whose runtime closure (the 9 distributions of ADR 0069 D3) is not at the packaged lock ead4edf1… records no dependency-set identity: its lock_sha256 is empty, every replay of it is inspected_archived_results / NOT_RUN, and the replaying side's failing distributions are named. Optional extras are not part of the identity.

`L42`: L42¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: pr-c1-v1 (Peng–Robinson with the light gases vapour-only, ADR 0026), shipped since M01, is bound by the C1 models (c1.*) on their own basis (R-288) and does not share the axes' components and domain: it takes the five real components H2, N2, NH3, Ar, CH4 (the only liquid is pure NH3) over 200–1000 K and 10⁴–3 × 10⁷ Pa. It is validated for pure-component behaviour only, within M01 spec §11's bands (NH3 saturation pressure within 2 %, liquid NH3 and light-gas ln φ within 0.05 of the reference equations of state). There is no mixture VLE validation, and k_ij = 0 for every pair: k_H2–NH3 = 0.1 would move the separator's vapour NH3 fraction y* by −3.1 % (about 3 % per 0.1; spec §4.2).

`L43`: L43¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: The Anderson recycle, its stagnation restart (recovery edge E1) and the stalled-recycle EO merge (E2) are library capabilities: every shipped unit that can sit in a loop declares exact derivatives, so auto resolves to newton_tear, and no application-offered policy requests anderson (docs/recovery-edges.yaml, enabled: library-only, R-130). V13 (b)'s and V14 (c)'s evidence for them is library-level.

`L44`: L44¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: The C1 reactor (c1.reactor) is measured on Linux x86-64 only, on one machine (ref-x86-64). Its runs are opt-in: they need the separately built, pinned reactor environment, which no CI job builds (M02 N2). They are reproducibility class R3 (ADR 0034 D6). A run on aarch64 is not measured, nor is the reproduction of a coupled run across architectures. macOS is best effort, and Windows is refused before spawn (ADR 0033 D2).

`L45`: L45¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: external-subprocess-v1 runs the group's pinned code with the user's privileges and is not a sandbox: there is no filesystem confinement, network isolation, memory or CPU quota, or syscall filtering, and trust rests on provenance (ADR 0033 D2; M02 N1).

`L46`: L46¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: No reactor result has been compared with experimental data by this project. The group's published comparison (Gargiulo et al. 2025 §3.1, of Rossetti et al. 2006's rate law) is the group's evidence. Every real-reactor number is the group's model at the pin, a regression, and its discretization bias is 1.0–1.4 % in NH₃ (M02 design note §13).

`L47`: L47¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: pr-c1-v1's saturated-liquid NH₃ volume is 11–30 % high (240–400 K) and its enthalpy of vaporization 2–5 % high (−7 % at 400 K), against the reference equations of state. No v0.2 loop decision uses either.

`L48`: L48¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: The C1 loop's convergence is evidenced at the corners and the centre of its registered hard domain only (M02 G11). The inert-fraction floor y_inert ≥ 0.035 sits just above a measured stage-1 failure (0.03 at 653.15 K, 11 MPa, H₂/N₂ 2.5). A failure inside the domain ends in a typed refusal and then COUPLING_NOT_CONVERGED, never in a false success.

`L49`: L49¶

Kind: Notice: not a failure.

No plain-language explanation is written yet; the summary is the technical one.

Summary: No promotable surrogate of the real C1 reactor exists in v0.2. M04's one real iteration (plan b2-it1 on box v2, reactor variant v3) is NOT_PROMOTABLE (band_not_finite): 26 of its 632 registered experiments, 5 of them among the 118 calibration draws, end in a deterministic reactor_not_accepted(certificate), more than the n − k = 4 refused draws the conformal band tolerates, so q̂ is infinite. A revision binds c1.reactor_surrogate only through a manifest that re-derives PROMOTABLE (R-360), and no further iteration on v3 is admissible (R-364).

Limitations¶

L-COL-1: L-COL-1¶

L-COL-10: L-COL-10¶

L-COL-2: L-COL-2¶

L-COL-3: L-COL-3¶

L-COL-4: L-COL-4¶

L-COL-5: L-COL-5¶

L-COL-6: L-COL-6¶

L-COL-7: L-COL-7¶

L-COL-8: L-COL-8¶

L-COL-9: L-COL-9¶

L-CSTR-1: L-CSTR-1¶

L-CSTR-2: L-CSTR-2¶

L-CSTR-3: L-CSTR-3¶

L-E01-1: L-E01-1¶

L-GP-1: L-GP-1¶

L-GP-2: L-GP-2¶

L-GP-3: L-GP-3¶

L-OPT-1: L-OPT-1¶

L-OPT-2: L-OPT-2¶

L-S11-1: L-S11-1¶

L-S11-2: L-S11-2¶

L-S11-3: L-S11-3¶

L-TH-1: L-TH-1¶

L-TH-2: L-TH-2¶

L-TH-3: L-TH-3¶

L-TH-4: L-TH-4¶

L-TH-5: L-TH-5¶

L-TH-6: L-TH-6¶

L-TH-7: L-TH-7¶

L-TH-8: L-TH-8¶

L-WEB-1: L-WEB-1¶

L-WEB-2: L-WEB-2¶

L-WEB-3: L-WEB-3¶

L-WEB-4: L-WEB-4¶

L-WS-1: L-WS-1¶

L-WS-2: L-WS-2¶

L-X04-1: L-X04-1¶

L-X04-2: L-X04-2¶

L-X04-3: L-X04-3¶

L01: L01¶

L02: L02¶

L03: L03¶

L04: L04¶

L05: L05¶

L06: L06¶

L07: L07¶

L08: L08¶

L09: L09¶

L10: L10¶

L11: L11¶

L12: L12¶

L13: L13¶

L14: L14¶

L15: L15¶

L16: L16¶

L17: L17¶

L18: L18¶

L19: L19¶

L20: L20¶

L21: L21¶

L22: L22¶

L23: L23¶

L24: L24¶

L25: L25¶

L26: L26¶

L27: L27¶

L28: L28¶

L29: L29¶

L30: L30¶

L31: L31¶

L32: L32¶

L33: L33¶

L34: L34¶

L35: L35¶

L36: L36¶

L37: L37¶

L38: L38¶

L39: L39¶

L40: L40¶

`L-COL-1`: L-COL-1¶

`L-COL-10`: L-COL-10¶

`L-COL-2`: L-COL-2¶

`L-COL-3`: L-COL-3¶

`L-COL-4`: L-COL-4¶

`L-COL-5`: L-COL-5¶

`L-COL-6`: L-COL-6¶

`L-COL-7`: L-COL-7¶

`L-COL-8`: L-COL-8¶

`L-COL-9`: L-COL-9¶

`L-CSTR-1`: L-CSTR-1¶

`L-CSTR-2`: L-CSTR-2¶

`L-CSTR-3`: L-CSTR-3¶

`L-E01-1`: L-E01-1¶

`L-GP-1`: L-GP-1¶

`L-GP-2`: L-GP-2¶

`L-GP-3`: L-GP-3¶

`L-OPT-1`: L-OPT-1¶

`L-OPT-2`: L-OPT-2¶

`L-S11-1`: L-S11-1¶

`L-S11-2`: L-S11-2¶

`L-S11-3`: L-S11-3¶

`L-TH-1`: L-TH-1¶

`L-TH-2`: L-TH-2¶

`L-TH-3`: L-TH-3¶

`L-TH-4`: L-TH-4¶

`L-TH-5`: L-TH-5¶

`L-TH-6`: L-TH-6¶

`L-TH-7`: L-TH-7¶

`L-TH-8`: L-TH-8¶

`L-WEB-1`: L-WEB-1¶

`L-WEB-2`: L-WEB-2¶

`L-WEB-3`: L-WEB-3¶

`L-WEB-4`: L-WEB-4¶

`L-WS-1`: L-WS-1¶

`L-WS-2`: L-WS-2¶

`L-X04-1`: L-X04-1¶

`L-X04-2`: L-X04-2¶

`L-X04-3`: L-X04-3¶

`L01`: L01¶

`L02`: L02¶

`L03`: L03¶

`L04`: L04¶

`L05`: L05¶

`L06`: L06¶

`L07`: L07¶

`L08`: L08¶

`L09`: L09¶

`L10`: L10¶

`L11`: L11¶

`L12`: L12¶

`L13`: L13¶

`L14`: L14¶

`L15`: L15¶

`L16`: L16¶

`L17`: L17¶

`L18`: L18¶

`L19`: L19¶

`L20`: L20¶

`L21`: L21¶

`L22`: L22¶

`L23`: L23¶

`L24`: L24¶

`L25`: L25¶

`L26`: L26¶

`L27`: L27¶

`L28`: L28¶

`L29`: L29¶

`L30`: L30¶

`L31`: L31¶

`L32`: L32¶

`L33`: L33¶

`L34`: L34¶

`L35`: L35¶

`L36`: L36¶

`L37`: L37¶

`L38`: L38¶

`L39`: L39¶

`L40`: L40¶

`L41`: L41¶