Revised review reflects:
- #1, #4 verified as non-issues (correct by design)
- #3, #5, #6, #8, #13 addressed in commit e40d6b4
- Updated recommendation to approve and merge
- Remaining items are low-priority style suggestions for future PRs

- P1 #1/#2: Add _validate_group_constant_strata_psu() helper and call
  it from fit() after the weight_type/replicate-weights checks. The
  dCDH IF expansion psi_i = U[g] * (w_i / W_g) treats each group as
  the effective sampling unit; when strata or PSU vary within group it
  silently spreads horizon-specific IF mass across observations in
  different PSUs, contaminating the stratified-PSU variance. Walk back
  the overstated claim at the old line 669 comment to match. Within-
  group-varying weights remain supported.
- P1 #3: _survey_se_from_group_if now filters zero-weight rows before
  np.unique/np.bincount so NaN / non-comparable group IDs on excluded
  subpopulation rows cannot crash SE factorization. psi stays full-
  length with zeros in excluded positions to preserve alignment with
  resolved.strata / resolved.psu inside compute_survey_if_variance.
- REGISTRY.md line 652 Note updated: explicitly states the
  within-group-constant strata/PSU requirement and the
  within-group-varying weights support.
- Tests: new TestSurveyWithinGroupValidation class (4 tests — rejects
  varying PSU, rejects varying strata, accepts varying weights, and
  ignores zero-weight rows during the constancy check) plus
  TestZeroWeightSubpopulation.test_zero_weight_row_with_nan_group_id.

All 268 targeted tests pass.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Addresses the second-round CI review findings:

- P1 false-pass (remaining): removed five phase-local try/except blocks
  that swallowed sub-step exceptions (HonestDiD M-grids in brand-awareness
  and BRFSS, dCDH HonestDiD and heterogeneity refit, dose-response
  dataframe extraction). Exceptions now escape, the phase is marked
  ok=false, and run_scenario's atexit handler exits nonzero. The fix
  caught a real API-usage bug on its first rerun: dose_response extract
  phase tried to pull event_study level on a result fit with
  aggregate="dose"; the event_study fit lives in a dedicated phase, so
  that level is removed from the extraction loop.
- P2 scenario-spec drift: BRFSS scenario text now says pweight TSL
  stage-2 (matching the aggregate_survey-returned design), not "Full
  replicate-weight path"; dCDH reversible scenario text now says
  heterogeneity="group" (matching the script), not "cohort".
- P3 path leakage: tracemalloc output now scrubs $HOME, repo root, and
  site-packages before writing the committed txt.

Drift-prevention layer:

- gen_findings_tables.py reads every JSON baseline and rewrites the
  numerical tables in performance-plan.md between
  <!-- TABLE:start <id> --> / <!-- TABLE:end <id> --> markers. Tables
  now re-derive from data on every rerun, eliminating the hand-edit
  drift the prior review flagged. Narrative prose stays hand-written
  by design, forcing a human re-read of findings when numbers shift.

Findings refresh (the numbers moved slightly; three narrative claims
needed updating):

- "Rust marginally slower than Python on JK1 at large scale" -> removed;
  fresh data has Rust and Python within noise on brand awareness at
  large (JK1 phase 0.577s Py / 0.562s Rust, totals 1.03 / 1.04).
- "ImputationDiD consistently dominant phase at all scales" -> narrowed
  to "dominant under Python; tied with SunAbraham under Rust at large".
- "Nine-figures of MB" in memory finding #3 was a phrasing error
  (literally 100+ TB); corrected to "mid-100s of MB".

Priority of optimization opportunities refreshed against new data:

- #1 aggregate_survey precompute stratum scaffolding: High (unchanged,
  now strongly supported - 24.75s Python / 25.41s Rust at 1M rows, 100%
  of chain runtime, growth only +31 MB).
- #2 Staggered CS working-memory audit: Low with explicit bump-trigger
  (Rust large crosses 512 MB Lambda line).
- #5 Rust-port JK1 replicate fit loop: demoted from Medium to Low -
  the "Rust regression to fix" leg of the rationale is gone because
  Rust is no longer slower.

Net: one clear priority (aggregate_survey fix), four optional follow-ups.
Still measurement only. No changes under diff_diff/ or rust/.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

P1 #1 — Stute tie-safe CvM:
Paper defines c_G(d) = Σ 1{D ≤ d} · eps with c_G(D_g) evaluated AT each
observation's dose, so tied observations share the post-tie cumulative sum.
My naive cumsum over sorted residuals produced partial within-tie sums that
were row-order-dependent. Fix: after cumsum, replace within-tie-block values
with the block's last cumsum via np.unique + np.repeat. `_cvm_statistic` now
accepts `d_sorted` and collapses tie blocks before squaring. Regression
test `test_cvm_statistic_tie_safe_order_invariance` pins order-invariance
on duplicate doses at atol=1e-14; `test_stute_order_invariance_with_duplicate_doses`
validates the end-to-end stute_test contract.

P1 #2 — Exact-linear fit must fail-to-reject (not return NaN):
For dy = a + b·d exact, Assumption 8 holds exactly and the correct outcome
is p=1, reject=False. My previous var(eps)<=0 check routed this to NaN. Fix:
dropped var(eps) degeneracy branch from stute_test (the bootstrap naturally
produces p=1 when eps=0 exactly). Added a scale-relative short-circuit
(sum(eps²) ≤ 1e-24 · sum(dy²)) in both stute_test and yatchew_hr_test so
FP noise (eps ~ 1e-16 from IEEE arithmetic on dy = 1 + 2*d) doesn't defeat
the short-circuit by producing non-zero but tiny OLS residuals. Yatchew
exact-linear now returns (t_stat_hr=-inf, p=1, reject=False) rather than
NaN. Regressions: TestStuteTest.test_exact_linear_returns_p1_not_nan,
TestYatchewHRTest.test_exact_linear_returns_p1_not_nan.

P1 #3 — HADPretestReport.all_pass contract:
Previously `all_pass = not (reject or reject or reject)` could be True
while `verdict` said "inconclusive - X NaN". Fix: gate all_pass on every
constituent p-value being finite AND no test rejecting. Updated docstring.
Regression: TestCompositeWorkflow.test_all_pass_false_when_any_test_nan.

P2 #1 — QUG negative-dose guard:
HAD doses must be non-negative (paper Section 2). The raw qug_test API
was silently folding d < 0 rows into the n_excluded_zero counter (filter
was `d > 0`). Fix: front-door ValueError on any d < 0. Regression:
TestQUGTest.test_negative_dose_raises.

P3 #1 — QUG np.partition:
REGISTRY claims O(G) via np.partition. Code was using np.sort. Switched
qug_test to np.partition(d_nz, 1), which guarantees partitioned[0] ≤
partitioned[1] = D_{(2)}, i.e., partitioned[0] = D_{(1)}. Tight
closed-form parity at atol=1e-12 still holds.

P3 #2 — REGISTRY n_bootstrap default:
REGISTRY said "Default n_bootstrap = 499" but code ships 999. Updated
REGISTRY to match code and added a note about the n_bootstrap >= 99
front-door validation.

Test count: 47 -> 53.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

dCDH by_path + placebo: per-path backward-horizon placebos (Wave 2 #3)

R5 was ✅ Looks good — only P3 polish remained. All addressed:

P3 #1 — exact-pin nprobust:
The parity contract runs through nprobust numerical paths
(DIDHAD's local-linear bandwidth + bias-correction calls), so a
fresh regeneration could drift if CRAN serves a newer nprobust.
Pin nprobust == 0.5.0 in both the R generator's stopifnot guard
and the parity test's metadata assertion alongside DIDHAD and
YatchewTest.

P3 #2 — workflow docstring:
did_had_pretest_workflow's top-level docstring still said "Eq 18
linear-trend detrending is a Phase 4 follow-up" which contradicts
the shipped trends_lin behavior. Updated to describe the
forwarding contract (trends_lin → joint_pretrends_test +
joint_homogeneity_test, consumed-placebo skip path on minimal
panels). Same fix on the StuteJointResult class docstring.

P3 #3 — parity test horizon-shape assertions:
Added an explicit "missing in Python" assertion in _zip_r_python:
every R-mapped event time must be present in Python's event_times
(catches future horizon-shape regressions where Python silently
drops a horizon R requested). Added an effects+placebo row-count
sanity check in test_yatchew_t_stat_parity (uses the previously-
unused effects/placebo parametrize values to catch fixture drift).

Stats: 540 tests pass, 0 regressions. No estimator/methodology
changes — all P3 polish.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

R5 was ✅ Looks good — only P3 polish remained. All addressed:

P3 #1 — exact-pin nprobust:
The parity contract runs through nprobust numerical paths
(DIDHAD's local-linear bandwidth + bias-correction calls), so a
fresh regeneration could drift if CRAN serves a newer nprobust.
Pin nprobust == 0.5.0 in both the R generator's stopifnot guard
and the parity test's metadata assertion alongside DIDHAD and
YatchewTest.

P3 #2 — workflow docstring:
did_had_pretest_workflow's top-level docstring still said "Eq 18
linear-trend detrending is a Phase 4 follow-up" which contradicts
the shipped trends_lin behavior. Updated to describe the
forwarding contract (trends_lin → joint_pretrends_test +
joint_homogeneity_test, consumed-placebo skip path on minimal
panels). Same fix on the StuteJointResult class docstring.

P3 #3 — parity test horizon-shape assertions:
Added an explicit "missing in Python" assertion in _zip_r_python:
every R-mapped event time must be present in Python's event_times
(catches future horizon-shape regressions where Python silently
drops a horizon R requested). Added an effects+placebo row-count
sanity check in test_yatchew_t_stat_parity (uses the previously-
unused effects/placebo parametrize values to catch fixture drift).

Stats: 540 tests pass, 0 regressions. No estimator/methodology
changes — all P3 polish.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

R1 P1 #1: pin remaining dismissal invariants
- Comment block claims 4 invariants hold but only invariants #1 (no
  execution) and #2 (fork-skip) had test coverage. Add 3 tests:
  - test_workflow_codex_step_uses_read_only_sandbox (invariant #1
    other half: sandbox: read-only)
  - test_workflow_resolve_pr_sets_head_sha_from_api (invariant #4:
    head_sha API-pinned, not from event payload)
  - test_workflow_comment_triggers_require_author_association
    (invariant #3: comment triggers gated on
    OWNER/MEMBER/COLLABORATOR)

R1 P1 #2: make guard test fail-closed across run scalar styles
- Prior regex only matched `run: |` literal blocks; inline `run: pytest`
  and folded `run: >` bypassed the scan entirely.
- Extract _extract_all_run_content static method that handles all three
  scalar styles (literal `|` with chomping variants, folded `>` with
  variants, and inline single-line). Both existing tests and a new
  python-file-exec test now use it.
- Expand FORBIDDEN_EXECUTION_PATTERNS to include `pip3 install` and
  `npm ci` (reviewer-named omissions).
- Add test_workflow_no_python_file_execution_against_workspace: regex
  flags `python(3)? <path>.py` invocations against workspace-relative
  paths (PR-head bytes), allowlists /tmp/-prefixed paths (BASE-staged
  via git show). Inline scripts (-c) and module invocations (-m) don't
  capture .py tokens, naturally excluded.

Test-the-test verified inline + folded + literal + npm ci + python
workspace all fire; python /tmp/ correctly does not. All 24 workflow
tests pass.

…prose

L194 checklist update (this PR) said the Eq. 18 detrending variant
shipped in PR #389; the explanatory prose immediately below at L200
still said "Phase 4 extends it with the Eq (18) detrending" as if it
were future work. Rewritten to past tense matching the L194 closure
and the REGISTRY § "Note (Phase 4 — Eq 17 / Eq 18 linear-trend
detrending shipped)" framing. Only the Pierce-Schott numerical
replication remains waived (REGISTRY Deviations Note #3).

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

Local Codex review on commit 79e0962 returned ✅ with 3 P3s (all
documentation/coverage, no actionable P0/P1). Per the test-coverage P3
upgrade rule (feedback_test_coverage_gap_treat_as_actionable.md),
addressing all three:

P3 igerber#1 (Code Quality): `_compute_cr2_bm_contrast_dof` was missing the
`ndim` validation that the parallel one-way `_compute_bm_dof_from_contrasts`
helper has, so a stray `(k,)` 1-D vector would die with a low-level
indexing error instead of a contract error. Added the same shape-tuple
check pattern (`if contrasts.ndim != 2 or contrasts.shape[0] != k`).

P3 igerber#2 (Docs): two stale doc surfaces post-feature-lift —
  - `estimators.py:68-71` base estimator docstring still said MPD did
    NOT support cluster + hc2_bm. Rewrote to describe the new
    cluster-aware contrast-DOF support and flag survey CR2-BM as the
    remaining gate.
  - `tests/test_linalg_hc2_bm.py` module banner still said clustered
    CR2 BM was "deferred to a follow-up". Updated to describe both the
    per-coefficient and the new compound-contrast DOF surfaces, and
    narrow the deferral to the weighted CR2-BM case only.

P3 igerber#3 (Tests): the new MPD test only asserted finite output, so a
regression that silently fell back to the shared n-k DOF would still
pass. Added `test_multi_period_cluster_hc2_bm_avg_att_uses_clubsandwich_dof`
which fits MPD on the new R `mpd_clustered_avg_att_dof` fixture and
recovers the implied Satterthwaite DOF by inverting
`avg_p_value = 2 * (1 - t.cdf(|avg_t_stat|, df))` via scipy.brentq. The
recovered DOF must match the R `Wald_test(test="HTZ")$df_denom` at
atol=1e-6. Also pins that the implied DOF is much smaller than the
n-k fallback (~39 here) — catches a regression to the shared df path.

All 254 tests in tests/test_linalg_hc2_bm.py + test_estimators_vcov_type.py
+ test_estimators.py pass; lint clean.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Add tests/test_methodology_had.py (6 classes, 34 tests) with paper-
equation-numbered Verified Components walk-through against de
Chaisemartin, Ciccia, D'Haultfoeuille & Knau (2026) arXiv:2405.04465v6
covering Equations 3 / 7 / 11 / 18 / 29 and Theorems 1 / 3 / 4 / 7:

- TestHADTheorem1Design1Prime: Eq. 3 Design 1' WAS recovery + N(0,1)
  coverage check at n_replicates=200, G=1000 with KS-stat <= 0.05 and
  empirical 95% coverage >= 0.90
- TestHADTheorem3MassPoint: Eq. 11 / Theorem 3 mass-point WAS_{d_lower}
  recovery + Wald-IV closed-form equivalence at atol=1e-9
- TestHADTheorem4QUG: Theorem 4 limit-law distributional match against
  closed-form F(t) = t/(1+t) at KS-stat <= 0.05, n_draws=5000, G=2000
- TestHADTheorem7YatchewHR: Eq. 29 standard-normal limit, paper-literal
  sigma2_diff = 1/(2G) normalization lock
- TestHADJointStute: Section 4.2 step 2 + 4.3 mean-independence variant
  H0 fail-to-reject + H1 reject under nonlinear DGP
- TestHADDeviations: equal-weighting invariance, sup-t bootstrap gating,
  staggered-timing fail-closed ValueError, safe_inference joint NaN

Add Assumption 5/6 non-testability documentation:
- HeterogeneousAdoptionDiD class docstring: new "Non-testable assumptions
  (paper Section 3.1.2)" Notes block citing Section 3.1.2 + cross-
  referencing the existing fit-time UserWarning at had.py:3372-3390
- qug_test / stute_test / yatchew_hr_test / did_had_pretest_workflow:
  "Scope (what this test does NOT cover)" clauses in Notes sections
  explicitly stating tests verify ADJACENT assumptions (4 / 7 / 8) and
  CANNOT test Assumptions 5 or 6

Close paper-review checklist L182-L194 + REGISTRY HAD Implementation
Checklist L2602-L2604: Phase 1a/1b/1c implementation closures (panel
validator, design paths, local-linear backend, bias-corrected CI),
staggered-timing fail-closed ValueError, zero-dose UserWarning filter,
Assumption 5/6 non-testability documentation. L2604 (covariates=
Theorem 6 NotImplementedError) remains [ ] with explicit TODO.md
cross-reference (currently a Python TypeError, fail-closed).

Waive Phase-4 validation-harness items igerber#1 (Pierce-Schott 2016 Figure 2)
+ igerber#2 (Table 1 coverage rates) with documented rationale: R parity at
atol=1e-8 in test_did_had_parity.py (3 DGPs x 5 method combos, bit-exact
via rtol=0) is a strictly stronger correctness anchor than coverage-rate
MC. Paper Section 5.2 itself self-acknowledges NP estimators too noisy
to be informative on the LBD-restricted PNTR panel.

REGISTRY HAD section gains a consolidated Deviations block (5 entries
with framing header distinguishing Notes igerber#1-igerber#2 = implementation choices
from Notes igerber#3-igerber#4 = waived validation-harness work from igerber#5 = Library
extension for staggered-timing fail-closed). Existing scattered Note
entries at L2313 (equal-weighting) and L2398 (sup-t gating) referenced
from the new block.

METHODOLOGY_REVIEW.md HAD row promoted In Progress -> Complete, detail
section rewritten with Verified Components / Test Coverage / Corrections
Made / Deviations / Outstanding Concerns structure mirroring the Bacon /
TripleDifference Complete-row layout.

TODO.md: existing Phase 4 Pierce-Schott row annotated with the 2026-05-20
waiver decision + rationale; new follow-up row for covariates= Theorem 6
NotImplementedError +Theorem 6 pointer (Low priority).

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

- P3 (Methodology): the promoted HAD materials described the Eq. 17/18
  `trends_lin=True` linear-trend-detrended variant as "deferred per Phase 4".
  This conflated TWO different things: (a) the FEATURE — which is shipped
  via the `trends_lin: bool = False` keyword-only kwarg on HAD.fit(),
  joint_pretrends_test, and joint_homogeneity_test (PR igerber#389; R-parity locked
  against DIDHAD::did_had(trends_lin=TRUE) v2.0.0 in test_did_had_parity.py);
  and (b) the PIERCE-SCHOTT NUMERICAL REPLICATION against the published
  p=0.51 anchor on the LBD-restricted panel, which IS waived per REGISTRY
  Deviations Note igerber#3. Updated 3 surfaces (paper-review L194, METHODOLOGY_REVIEW
  Eq. 18 Verified-Components row, test_methodology_had.py module docstring +
  TestHADJointStute class docstring) to distinguish "feature shipped + R-parity
  locked elsewhere" from "Pierce-Schott numerical replication waived".

- P3 (Documentation/Tests): TestHADJointStute promotion narrative overstated
  H1 coverage as "H0 fail-to-reject and H1 reject on linear vs nonlinear DGPs"
  for both joint_pretrends_test and joint_homogeneity_test. Reality: H1
  rejection is tested only on joint_homogeneity_test via a quadratic post-
  DGP; joint_pretrends_test gets H0-only coverage in this file (H1 would
  require a violating-pretrends fixture that re-verifies bootstrap calibration
  covered by test_had_pretests.py). Narrowed wording in METHODOLOGY_REVIEW
  Verified-Components row + TestHADJointStute class docstring; CHANGELOG entry
  unchanged (the H1 reject claim in CHANGELOG explicitly cites the homogeneity
  side via "H1 reject under nonlinear DGP", which is accurate).

All 35 methodology tests pass; lint clean.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

…GELOG H1 scope

- R6 fix left METHODOLOGY_REVIEW.md Deviations item igerber#6 stale (only updated
  the Verified-Components row). Item igerber#6 still said "Eq. 18 linear-trend-
  detrended joint Stute deferred". Rewritten to match the rest of the
  HAD tracker: trends_lin=True is SHIPPED + R-parity-locked in
  test_did_had_parity.py; the methodology-walkthrough file deliberately
  doesn't duplicate that coverage; the Pierce-Schott published-value
  numerical replication is what's waived (Deviations Note igerber#3).

- R6 narrowed the Verified-Components row + class docstring but missed the
  CHANGELOG bullet, which still claimed "joint Stute pre-trends + homogeneity
  H0 fail-to-reject + H1 reject under nonlinear DGP". Narrowed to:
  "H0 fail-to-reject on both surfaces and H1 reject for joint homogeneity
  under a nonlinear DGP" — matches the test file's actual scope.

All 35 methodology tests pass; lint clean.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

… to_dict()

ImputationDiDResults now exposes `vcov_type`, `cluster_name`, `n_clusters`
and a new `to_dict()` method (Phase 1b interstitial #3), but the shared
"Common Results Pattern for Staggered Estimators" section in llms-full.txt
still listed only `summary()`, `print_summary()`, and `to_dataframe()`.
Adds a variance-metadata table and threads `to_dict()` into the Methods
line so AI-agent guide consumers can discover the surface.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Phase 1b interstitial #3 for ImputationDiD. Mirrors the CallawaySantAnna
(PR #487) + TripleDifference (PR #488) template for IF-based estimators:
vcov_type is permanently narrow to {"hc1"} because the per-unit influence
function aggregation (Borusyak-Jaravel-Spiess 2024 Theorem 3) has no
single design matrix on which hat-matrix leverage or Bell-McCaffrey
Satterthwaite DOF can be defined.

Source surface:
- diff_diff/imputation.py: vcov_type param + @staticmethod
  _validate_vcov_type + fit()-time revalidation +
  cluster+replicate-weights NotImplementedError guard +
  Results cluster_name/n_clusters resolution
- diff_diff/imputation_results.py: vcov_type/cluster_name/n_clusters
  fields + new to_dict() + variance-estimator line in summary() routing
  through shared _format_vcov_label helper
- diff_diff/imputation_bootstrap.py: dual-site n_clusters<2 /
  n_psu<2 NaN guards via new _build_nan_bootstrap_results helper
  (closes the BLAS-roundoff zero-SE class predicted to recur on
  IF-based estimators)

Tests: 34 new tests in TestImputationDiDVcovType covering default /
cluster / TSL-survey / replicate-survey bit-equality (parametrized over
aggregate modes), bootstrap × cluster + bootstrap × survey bit-equality,
fit()-time revalidation after set_params bypass, bootstrap n_psu<2 /
n_clusters<2 NaN propagation, pretrends bit-equality, and the full
introspection + safety-gate surface (8 tests).

Docs: REGISTRY.md (IF-based taxonomy + 4 new Notes), CHANGELOG.md,
TODO.md (row narrowed, Conley follow-up added), llms-full.txt
(vcov_type + pretrends signature drift fix).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

… to_dict()

ImputationDiDResults now exposes `vcov_type`, `cluster_name`, `n_clusters`
and a new `to_dict()` method (Phase 1b interstitial #3), but the shared
"Common Results Pattern for Staggered Estimators" section in llms-full.txt
still listed only `summary()`, `print_summary()`, and `to_dataframe()`.
Adds a variance-metadata table and threads `to_dict()` into the Methods
line so AI-agent guide consumers can discover the surface.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

… 3 (Phase 1b interstitial #3)

Phase 1b interstitial #3 for ImputationDiD. Mirrors the CallawaySantAnna
(PR #487) + TripleDifference (PR #488) template for IF-based estimators:
vcov_type is permanently narrow to {"hc1"} because the per-unit influence
function aggregation (Borusyak-Jaravel-Spiess 2024 Theorem 3) has no
single design matrix on which hat-matrix leverage or Bell-McCaffrey
Satterthwaite DOF can be defined.

Source surface (diff_diff/):
- imputation.py: vcov_type param + @staticmethod _validate_vcov_type +
  fit()-time revalidation + cluster+replicate-weights NotImplementedError
  guard + Results metadata resolution (cluster_name=unit by default for
  the Theorem 3 unit-clustered IF variance; suppressed under ANY survey
  design — analytical OR replicate — because replicate variance ignores
  cluster/PSU entirely)
- imputation_results.py: vcov_type/cluster_name/n_clusters fields, new
  to_dict() method, summary() variance line via shared _format_vcov_label
  (default cluster=None renders "CR1 cluster-robust at <unit>, G=<n>";
  bootstrap fits suppress the analytical label and render
  "Inference method: bootstrap" instead, mirroring DiDResults.summary()
  gate at results.py:213-226)
- imputation_bootstrap.py: dual-site n_clusters<2 / n_psu<2 NaN guards
  via new _build_nan_bootstrap_results helper (closes the BLAS-roundoff
  zero-SE class predicted to recur on IF-based estimators)

Tests: 42 new tests in TestImputationDiDVcovType covering default /
cluster / TSL-survey / replicate-survey + bootstrap × cluster + bootstrap
× survey bit-equality (ALL parametrized over aggregate ∈ {None,
"event_study", "group"} with per-horizon and per-group SE override
branches pinned); fit()-time revalidation after set_params bypass;
bootstrap n_psu<2 / n_clusters<2 NaN propagation including coef_var NaN;
pretrends=True × vcov_type='hc1' × cluster bit-equality; introspection
(default attr, get_params, Results carries, to_dict, summary label
default+cluster+bootstrap-suppressed, cluster_name suppression under
both analytical AND replicate survey, fit-clone idempotence,
convenience function); input rejection on classical/hc2/hc2_bm/conley/
unknown with distinct methodology-keyword pins; cluster+replicate
rejection. Full pytest tests/test_imputation.py: 128 passed.

Docs:
- REGISTRY.md: IF-based taxonomy adds ImputationDiD to "Enforced today"
  tier; ImputationDiD section gains 4 new Notes (vcov_type contract,
  cluster+replicate fail-closed, bootstrap n<2 NaN, default unit-cluster
  CR1 rendering)
- CHANGELOG.md: [Unreleased] entry
- TODO.md: Phase 1b row narrowed to TwoStageDiD + EfficientDiD;
  ImputationDiD Conley follow-up row added
- guides/llms-full.txt: vcov_type + pretrends signature drift fix +
  shared staggered-results section advertises new variance metadata
  fields and to_dict()

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…-phase1b

ImputationDiD: thread vcov_type as narrow {hc1} contract per BJS Theorem 3 (Phase 1b interstitial #3)