From 4c437932813fc41850b7c5f247b1a0da3b811e2b Mon Sep 17 00:00:00 2001
From: Claude <noreply@anthropic.com>
Date: Thu, 29 Oct 2026 21:50:55 +0000
Subject: [PATCH] Carry distance_metadata and global-clock oneepoch whole as v1

[reshaped retire classes carried] (Jess Haley, 2026-10-29, did-schema
V_eta_go_forward_class_audit.md, "B"): v1 distance_metadata and oneepoch
are carried whole as v1 for now, not reshaped into their retire classes.

- +migrators_j/oneepoch.m: a oneepoch on a global clock, which kept the
  class with its element_epoch block folded on, is carried as the untouched
  v1 source citing carry-148 (awaiting_home). The carry-126 (device-local)
  and carry-100 branches are unchanged; the fold code is removed.
- +migrators_j/distance_metadata.m: every document the reshape reached is
  carried citing carry-147 (awaiting_home); carry-100/carry-049 unchanged;
  the endpoint-reshape helpers are removed.
- The two round-trip inverses are kept, noted as reached by no migrator
  output (as position_metadata's). Contents.m roster regenerated.

Tests inverted, not deleted: testMigratorsJ (distance_metadata and the
four oneepoch tests), testRawV1RoundTripOneEpoch, testRawV1RoundTripPosition.

Needs did-schema #238 (register rows carry-147, carry-148) on main.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FxKLA8Tbf9cpvDe7fuVeKf
---
 src/did/+did2/+convert/+migrators/Contents.m  |   5 +-
 .../+did2/+convert/+migrators_j/Contents.m    |   5 +-
 .../+convert/+migrators_j/distance_metadata.m | 150 ++---------
 .../+convert/+migrators_j/element_epoch.m     |   4 +-
 .../+did2/+convert/+migrators_j/oneepoch.m    |  90 ++++---
 .../+convert/+roundtrip/distance_metadata.m   |   7 +
 src/did/+did2/+convert/+roundtrip/oneepoch.m  |   7 +
 tests/+did2/+unittest/testFixtureCorpus.m     |   9 +-
 tests/+did2/+unittest/testMigrators.m         |   3 +-
 tests/+did2/+unittest/testMigratorsJ.m        | 250 +++++++-----------
 .../+unittest/testRawV1RoundTripOneEpoch.m    |  52 ++--
 .../+unittest/testRawV1RoundTripPosition.m    |  62 +++--
 .../+unittest/testResolveEpochProbemap.m      |   3 +-
 13 files changed, 283 insertions(+), 364 deletions(-)

diff --git a/src/did/+did2/+convert/+migrators/Contents.m b/src/did/+did2/+convert/+migrators/Contents.m
index 2e8548cc..e83c0009 100644
--- a/src/did/+did2/+convert/+migrators/Contents.m
+++ b/src/did/+did2/+convert/+migrators/Contents.m
@@ -42,8 +42,9 @@
 %
 %   The position_metadata, distance_metadata and stimulus_bath base migrators
 %   are deleted too: each class has a +migrators_j migrator, which reads raw v1
-%   (distance_metadata's reshapes the v1 spelling itself, where it used to hand
-%   the renamed body to the base migrator).
+%   (distance_metadata's reshaped the v1 spelling itself, where it used to hand
+%   the renamed body to the base migrator; since [reshaped retire classes
+%   carried], Jess Haley, 2026-10-29, it carries every document whole as v1).
 %
 %   The element, ontology_image and ontology_label base migrators are deleted
 %   (2026-10-27): each class has a +migrators_j migrator that reads raw v1, and
diff --git a/src/did/+did2/+convert/+migrators_j/Contents.m b/src/did/+did2/+convert/+migrators_j/Contents.m
index c4da6df5..dfea9d20 100644
--- a/src/did/+did2/+convert/+migrators_j/Contents.m
+++ b/src/did/+did2/+convert/+migrators_j/Contents.m
@@ -695,8 +695,7 @@
 %         deleted 2026-10-21; the class is now `schemas/V_eta/stable/demo.json`); the
 %         fold itself is in private/jDemoFold.
 %     distance_metadata
-%         Brainstorm-J migrator: reshape a did_v1 distance_metadata into the validated
-%         V_eta distance_metadata. 1 -> 1.
+%         Brainstorm-J migrator: carry a v1 distance_metadata whole as v1. 1 -> 1.
 %     electrode_offset_voltage
 %         Brainstorm-J migrator: did_v1 electrode_offset_voltage -> an `observation`
 %         with value kind `voltage` about the probe's entity, qualified by the
@@ -820,7 +819,7 @@
 %         sorted unit's own entity (+ the shared session anchor, the software and its
 %         run); no entity of its own.
 %     oneepoch
-%         Brainstorm-J migrator: keep the class, fold its inherited block.
+%         Brainstorm-J migrator: carry a v1 oneepoch whole as v1. 1 -> 1.
 %     ontology_image
 %         Brainstorm-J migrator: did_v1 ontology_image, dispatched ON SHAPE. The
 %         current NDI shape (`ontology_nodes` + an `ontologyTableRow_id` edge + the
diff --git a/src/did/+did2/+convert/+migrators_j/distance_metadata.m b/src/did/+did2/+convert/+migrators_j/distance_metadata.m
index 676d693f..e547561b 100644
--- a/src/did/+did2/+convert/+migrators_j/distance_metadata.m
+++ b/src/did/+did2/+convert/+migrators_j/distance_metadata.m
@@ -1,6 +1,22 @@
 function bodies = distance_metadata(preBody)
-%DISTANCE_METADATA Brainstorm-J migrator: reshape a did_v1 distance_metadata into
-%   the validated V_eta distance_metadata. 1 -> 1.
+%DISTANCE_METADATA Brainstorm-J migrator: carry a v1 distance_metadata whole as
+%   v1. 1 -> 1.
+%
+%   ================ [reshaped retire classes carried] =====================
+%   Decided by Jess Haley, 2026-10-29 (did-schema
+%   V_eta_go_forward_class_audit.md, "B"): v1 `distance_metadata` and
+%   `oneepoch` are carried whole as v1 for now, not reshaped into their retire
+%   classes; the reshape waits for their V2 homes (`distance_metadata`: a
+%   decision still to make, docs/V2_PLAN.md G50). So EVERY distance_metadata
+%   is carried whole as v1 (the untouched v1 source, with a `carried_as_v1`
+%   note): one holding a field outside its NDI template under carry-100, as
+%   before, and every other one under register row carry-147, awaiting_home.
+%   The reshape into the V_eta `distance_metadata` (a class whose disposition
+%   is `retire`) described below is no longer done; the description is the
+%   record of what was replaced. did2.convert.roundtrip.distance_metadata, its
+%   inverse, is reached by no migrator output any more; a carried document's
+%   round trip is the carry itself (did2.convert.roundtrip.check).
+%   =======================================================================
 %
 %   Routed from did2.convert.v1_to_v2 only when TargetVersion == 'V_eta'. The v1
 %   doc is FLAT and records the two ENDPOINTS of a distance measurement, not the
@@ -96,34 +112,17 @@
         'distance_metadata body is missing the distance_metadata property block.');
 end
 
-% the endpoint labels: the v1 suffixes, and on the renamed path (v1_to_v2
-% 'RawV1Classes', {}, until universalRenames is deleted) the suffixes of the
-% spelling that pass gives them (ontology_node_a, integer_i_ds_a, ...)
-labels = upper(suffixes);
-renamed = renamedSuffixes(blk);
-labels = [labels, upper(renamed(~ismember(upper(renamed), labels)))];
-[labels, order] = sort(labels);
-raw = [suffixes, repmat({''}, 1, numel(labels) - numel(suffixes))];
-raw = raw(order);
-endpoints = struct('label', {}, 'measurement', {}, ...
-    'integer_ids', {}, 'string_ids', {}, 'numeric_values', {});
-for k = 1:numel(labels)
-    endpoints(end+1) = buildEndpoint(blk, labels{k}, raw{k}); %#ok<AGROW>
-end
-
-unitsNode = '';
-if isfield(blk, 'units')
-    unitsNode = char(blk.units);
-end
-
-v2Body = preBody;
-if isfield(v2Body, 'depends_on') && ~isempty(v2Body.depends_on)
-    % the {name, document_id} entries the renamed path gave the document
-    deps = did2.convert.v1read.readDependsOn(src);
-    v2Body.depends_on = struct('name', {deps.name}, 'document_id', {deps.document_id});
-end
-v2Body.distance_metadata = struct('endpoints', endpoints, 'units', ontologyTerm(unitsNode));
-bodies = {v2Body};
+% [reshaped retire classes carried] (Jess Haley, 2026-10-29, did-schema
+% V_eta_go_forward_class_audit.md): no longer reshaped into the retire class
+% `distance_metadata`; carried whole as v1 until its V2 home is decided.
+b = did2.convert.v1read.readBase(src);
+id = b.id;
+if ~ischar(id); id = ''; end
+% carry-register: carry-147
+did2.convert.reportNote('carried_as_v1', 'distance_metadata', id, ...
+    ['[carry-147] awaiting_home: distance_metadata, until its V2 home is ' ...
+    'decided ([reshaped retire classes carried])']);
+bodies = {src};
 end
 
 function suffixes = endpointSuffixes(block)
@@ -138,94 +137,3 @@
     end
 end
 end
-
-function suffixes = renamedSuffixes(block)
-% Every x with an `ontology_node_x` field (universalRenames' spelling).
-suffixes = {};
-fns = fieldnames(block);
-for k = 1:numel(fns)
-    tokens = regexp(fns{k}, '^ontology_node_(.+)$', 'tokens', 'once');
-    if ~isempty(tokens) && ~any(strcmpi(suffixes, tokens{1}))
-        suffixes{end+1} = tokens{1}; %#ok<AGROW>
-    end
-end
-end
-
-function [v, found] = endpointField(block, rawName, x, renamedName, label)
-% The v1 field <rawName>_<x>, else universalRenames' <renamedName>_<label>.
-v = [];
-found = false;
-names = {};
-if ~isempty(x)
-    names{end+1} = [rawName '_' x];
-end
-names{end+1} = [renamedName '_' lower(label)];
-for k = 1:numel(names)
-    if isfield(block, names{k})
-        v = block.(names{k});
-        found = true;
-        return;
-    end
-end
-end
-
-function ep = buildEndpoint(block, label, x)
-% One V2 endpoint record from v1's per-suffix fields (X, as the document
-% spells it, or '' when only the renamed spelling holds the endpoint). A field
-% the document does not have is an empty value of the right type.
-nodeVal = '';
-[v, found] = endpointField(block, 'ontologyNode', x, 'ontology_node', label);
-if found
-    nodeVal = char(v);
-end
-intIds = [];
-[v, found] = endpointField(block, 'integerIDs', x, 'integer_i_ds', label);
-if found
-    intIds = double(v);
-    intIds = intIds(:)';
-end
-stringIds = string.empty(1, 0);
-[raw, found] = endpointField(block, 'ontologyStringValues', x, 'ontology_string_values', label);
-if found
-    if ischar(raw)
-        parts = strsplit(raw, ',');
-        parts = cellfun(@strtrim, parts, 'UniformOutput', false);
-        parts = parts(~cellfun('isempty', parts));
-        stringIds = string(parts);
-    elseif isstring(raw)
-        stringIds = raw(:)';
-    end
-end
-nums = [];
-[v, found] = endpointField(block, 'ontologyNumericValues', x, 'ontology_numeric_values', label);
-if found
-    if isnumeric(v)
-        nums = double(v(:)');
-    end
-end
-ep = struct( ...
-    'label',          label, ...
-    'measurement',    ontologyTerm(nodeVal), ...
-    'integer_ids',    intIds, ...
-    'string_ids',     stringIds, ...
-    'numeric_values', nums);
-end
-
-function term = ontologyTerm(curie)
-% An ontology_term composite. Empty input yields a node-only, name-empty
-% struct. ndi.ontology.lookup failures leave name ''.
-term = struct('node', char(curie), 'name', '');
-if isempty(curie)
-    return;
-end
-try
-    [~, resolved] = ndi.ontology.lookup(char(curie));
-    if isstring(resolved) && isscalar(resolved)
-        term.name = char(resolved);
-    elseif ischar(resolved)
-        term.name = resolved;
-    end
-catch
-    % ndi.ontology not on path or curie unknown; leave name ''.
-end
-end
diff --git a/src/did/+did2/+convert/+migrators_j/element_epoch.m b/src/did/+did2/+convert/+migrators_j/element_epoch.m
index 07f9f099..0d8f02f7 100644
--- a/src/did/+did2/+convert/+migrators_j/element_epoch.m
+++ b/src/did/+did2/+convert/+migrators_j/element_epoch.m
@@ -102,8 +102,8 @@
 %       referent "the element_epoch doc and its binary"). When this dissolves the
 %       edge retargets to whatever absorbs the payload.
 %     * `oneepoch` inherits its whole shape from `element_epoch` and is handled by
-%       its own migrator as a fold-into-tombstone; fork A1 for it is chosen but
-%       NOT SIGNED, and is gated on #30 besides.
+%       its own migrator, which carries it whole as v1 (carry-126 / carry-148,
+%       [reshaped retire classes carried], Jess Haley, 2026-10-29).
 %
 %   ALSO STILL TRUE, and repairs rather than decisions:
 %   `acquisition_epoch.axes` / `.channels` / `.storage` appear in NO NDI template
diff --git a/src/did/+did2/+convert/+migrators_j/oneepoch.m b/src/did/+did2/+convert/+migrators_j/oneepoch.m
index afb2f520..2c5855a3 100644
--- a/src/did/+did2/+convert/+migrators_j/oneepoch.m
+++ b/src/did/+did2/+convert/+migrators_j/oneepoch.m
@@ -1,5 +1,12 @@
 function v2Body = oneepoch(preBody)
-%ONEEPOCH Brainstorm-J migrator: keep the class, fold its inherited block.
+%ONEEPOCH Brainstorm-J migrator: carry a v1 oneepoch whole as v1. 1 -> 1.
+%
+%   EVERY oneepoch IS CARRIED WHOLE AS v1 (the untouched v1 source, with a
+%   `carried_as_v1` note): one stitched on device-local time under carry-126,
+%   one on a global clock under carry-148, see THE 2026-10-27 / 2026-10-29
+%   DECISIONS below. The class is no longer kept and its inherited block is no
+%   longer folded; the sections up to READS RAW v1 are the record of what that
+%   fold did, kept until the global-clock calculation replaces it.
 %
 %   `oneepoch` is NOT an epoch class. It is the record of a CONCATENATION:
 %   `ndi.element.oneepoch` glues an element's N epochs into a single epoch, and
@@ -87,21 +94,32 @@
 %         +element/oneepoch.m: "as though the recordings occurred one
 %         immediately after the other") is carried whole as v1
 %         `needs_mapping`.
-%   BUILT HERE: the device-local carry. ndi.element.oneepoch picks the first
-%   GLOBAL clock of the input's epoch table (ndi.time.clocktype.isGlobal:
-%   utc, approx_utc, exp_global_time, approx_exp_global_time,
-%   dev_global_time, approx_dev_global_time) and stitches on
-%   dev_local_time only when there is none; it writes EVERY clock of the
-%   element, comma-joined, as element_epoch.epoch_clock. So a oneepoch none
-%   of whose clocks is global was stitched on device-local time, and is
-%   carried whole as v1 (isStitchedOnLocalTime), citing register row
-%   carry-126 (did-schema 64ad58b8).
-%   NOT BUILT YET: the global-clock calculation. It still keeps the class
-%   `oneepoch` as below, until the questions in the PR that builds the
-%   device-local carry are answered (which statements are "the per-epoch
-%   recordings" when the input's per-epoch observations are not in the
-%   batch; the calculation's variable and anchor; a sampled_body from a
-%   .vhsb the DID pass cannot read).
+%     [reshaped retire classes carried]  (Jess Haley, 2026-10-29, "B")
+%         v1 `distance_metadata` and `oneepoch` are carried whole as v1 for
+%         now, not reshaped into their retire classes; the reshape waits for
+%         their V2 homes (`oneepoch`: docs/V2_PLAN.md G44).
+%   BUILT HERE: both carries. ndi.element.oneepoch picks the first GLOBAL
+%   clock of the input's epoch table (ndi.time.clocktype.isGlobal: utc,
+%   approx_utc, exp_global_time, approx_exp_global_time, dev_global_time,
+%   approx_dev_global_time) and stitches on dev_local_time only when there is
+%   none; it writes EVERY clock of the element, comma-joined, as
+%   element_epoch.epoch_clock. So a oneepoch none of whose clocks is global
+%   was stitched on device-local time, and is carried whole as v1,
+%   needs_mapping (isStitchedOnLocalTime), citing register row carry-126
+%   (did-schema 64ad58b8). One with a global clock, which until 2026-10-29
+%   kept the class `oneepoch` with its element_epoch block folded onto it, is
+%   carried whole as v1, awaiting_home, citing register row carry-148
+%   ([reshaped retire classes carried]).
+%   NOT BUILT YET: the global-clock calculation and the batch pass for its
+%   per-epoch `input_id` links (docs/V2_PLAN.md G44); when it is built it
+%   replaces the carry-148 branch. Open questions: which statements are "the
+%   per-epoch recordings" when the input's per-epoch observations are not in
+%   the batch; the calculation's variable and anchor; a sampled_body from a
+%   .vhsb the DID pass cannot read.
+%
+%   did2.convert.roundtrip.oneepoch, the inverse of the kept class, is reached
+%   by no migrator output while every oneepoch is carried; a carry's round
+%   trip is the carry itself (did2.convert.roundtrip.check).
 
 arguments
     preBody (1,1) struct
@@ -149,11 +167,11 @@
 % V_eta_go_forward_class_audit.md: a oneepoch NDI stitched on device-local time
 % (none of its clocks global) is carried whole as v1, needs_mapping. Its
 % times exist in no input.
+id = '';
+if isfield(src, 'base') && isstruct(src.base) && isfield(src.base, 'id')
+    id = char(src.base.id);
+end
 if isStitchedOnLocalTime(v2Body.element_epoch.clocks)
-    id = '';
-    if isfield(src, 'base') && isstruct(src.base) && isfield(src.base, 'id')
-        id = char(src.base.id);
-    end
     % carry-register: carry-126
     did2.convert.reportNote('carried_as_v1', 'oneepoch', id, ...
         ['[carry-126] needs_mapping: a oneepoch stitched on ' ...
@@ -163,29 +181,15 @@
     return;
 end
 
-% The concrete class is unchanged -- `oneepoch` persists as a source tombstone
-% until the global-clock calculation ([oneepoch is a redundant calculation]) is
-% built; see THE 2026-10-27 / 2026-10-29 DECISIONS above.
-v2Body.document_class.class_name = 'oneepoch';
-
-if ~isfield(v2Body, 'oneepoch') || ~isstruct(v2Body.oneepoch)
-    v2Body.oneepoch = struct();
-end
-
-% Fold the inherited block onto the concrete one. Field-by-field rather than
-% wholesale, so a v1 field this migrator has never seen lands somewhere the
-% strict-fields check will REPORT it (`undeclaredField` on `oneepoch`) instead of
-% being carried in a block nothing validates.
-if isfield(v2Body, 'element_epoch')
-    inherited = v2Body.element_epoch;
-    if isstruct(inherited)
-        fns = fieldnames(inherited);
-        for k = 1:numel(fns)
-            v2Body.oneepoch.(fns{k}) = inherited.(fns{k});
-        end
-    end
-    v2Body = rmfield(v2Body, 'element_epoch');
-end
+% [reshaped retire classes carried] (Jess Haley, 2026-10-29, did-schema
+% V_eta_go_forward_class_audit.md): a oneepoch on a global clock, which kept
+% the class `oneepoch` until then, is carried whole as v1 until the
+% global-clock calculation ([oneepoch is a redundant calculation]) is built.
+% carry-register: carry-148
+did2.convert.reportNote('carried_as_v1', 'oneepoch', id, ...
+    ['[carry-148] awaiting_home: a oneepoch on a global clock, until its ' ...
+    'data-combination calculation is built ([reshaped retire classes carried])']);
+v2Body = {src};
 end
 
 function tf = isStitchedOnLocalTime(clocks)
diff --git a/src/did/+did2/+convert/+roundtrip/distance_metadata.m b/src/did/+did2/+convert/+roundtrip/distance_metadata.m
index f5ff3bc7..fa454bcc 100644
--- a/src/did/+did2/+convert/+roundtrip/distance_metadata.m
+++ b/src/did/+did2/+convert/+roundtrip/distance_metadata.m
@@ -25,6 +25,13 @@
 %
 %   Any other emitted document is unplaced.
 %
+%   NO MIGRATOR OUTPUT REACHES THIS INVERSE since [reshaped retire classes
+%   carried] (Jess Haley, 2026-10-29, did-schema
+%   V_eta_go_forward_class_audit.md): migrators_j.distance_metadata carries every document
+%   whole as v1 (carry-147), and a carry's round trip is the carry itself
+%   (did2.convert.roundtrip.check). It is kept until the class's V2 home is
+%   built, whose inverse replaces it.
+%
 %   See also did2.convert.roundtrip.check,
 %            did2.convert.migrators_j.distance_metadata.
 
diff --git a/src/did/+did2/+convert/+roundtrip/oneepoch.m b/src/did/+did2/+convert/+roundtrip/oneepoch.m
index a4e4e511..b92d0e9c 100644
--- a/src/did/+did2/+convert/+roundtrip/oneepoch.m
+++ b/src/did/+did2/+convert/+roundtrip/oneepoch.m
@@ -16,6 +16,13 @@
 %     element_epoch.epoch_clock        the clocks' names, comma-joined
 %     element_epoch.t0_t1              the clocks' [t0; t1], one column each
 %
+%   NO MIGRATOR OUTPUT REACHES THIS INVERSE since [reshaped retire classes
+%   carried] (Jess Haley, 2026-10-29, did-schema
+%   V_eta_go_forward_class_audit.md): migrators_j.oneepoch carries every document
+%   whole as v1 (carry-126 / carry-148), and a carry's round trip is the carry itself
+%   (did2.convert.roundtrip.check). It is kept until the class's V2 home is
+%   built, whose inverse replaces it.
+%
 %   See also did2.convert.roundtrip.check,
 %            did2.convert.migrators_j.oneepoch,
 %            did2.convert.roundtrip.element_epoch.
diff --git a/tests/+did2/+unittest/testFixtureCorpus.m b/tests/+did2/+unittest/testFixtureCorpus.m
index 39feea7f..1b228307 100644
--- a/tests/+did2/+unittest/testFixtureCorpus.m
+++ b/tests/+did2/+unittest/testFixtureCorpus.m
@@ -1251,10 +1251,11 @@ function testACarriedCalculatorWithNoTombstoneIsCountedNotQuarantined(testCase)
 % ---- distance_metadata (element_id) -- REAL FLAT v1 shape ------------------
 % The writer emits FLAT per-endpoint fields (ontologyNode_A/_B, integerIDs_A/_B,
 % ontologyStringValues_A/_B, ontologyNumericValues_A/_B empty by design, units),
-% NOT a nested `endpoints` block. The migrator reshapes flat -> the nested
-% `endpoints` array the schema requires; this fixture is the real shape so the
-% reshape is exercised under validation (element_id -> the minted subject; the
-% endpoint node ids are field values, not orphan-checked depends_on edges).
+% NOT a nested `endpoints` block. Until 2026-10-29 the migrator reshaped flat ->
+% the nested `endpoints` array of the retire class; since [reshaped retire
+% classes carried] (Jess Haley, 2026-10-29) it carries the document whole as v1
+% (carry-147), and this fixture checks the carry does not quarantine under
+% validation.
 function batch = fx_distance_metadata()
 sub = subjDoc('dm_sub', 'elemDM');
 d = struct();
diff --git a/tests/+did2/+unittest/testMigrators.m b/tests/+did2/+unittest/testMigrators.m
index be190d75..9963cb96 100644
--- a/tests/+did2/+unittest/testMigrators.m
+++ b/tests/+did2/+unittest/testMigrators.m
@@ -317,6 +317,7 @@ function testTheUnreachedBaseMigratorsAreGone(testCase)
 % --- position_metadata, distance_metadata, stimulus_bath ---
 % Their +migrators base migrators are deleted ([early migration pieces fixed]
 % (1)): never reached (each class has a +migrators_j migrator), or reached only
-% through distance_metadata's own, which now reshapes the raw v1 spelling. The
+% through distance_metadata's own, which reshaped the raw v1 spelling itself
+% (and since [reshaped retire classes carried], 2026-10-29, carries it). The
 % classes read raw v1 (did2.convert.rawV1Classes); their round trips are
 % testRawV1RoundTripBath and testRawV1RoundTripPosition.
diff --git a/tests/+did2/+unittest/testMigratorsJ.m b/tests/+did2/+unittest/testMigratorsJ.m
index b831d6d9..f6216893 100644
--- a/tests/+did2/+unittest/testMigratorsJ.m
+++ b/tests/+did2/+unittest/testMigratorsJ.m
@@ -4084,15 +4084,18 @@ function testElementEpochRenamedToAcquisitionEpoch(testCase)
 verifyEqual(testCase, out.acquisition_epoch.clocks(1).t1, 930.35);
 end
 
-function testDistanceMetadataReshapesFlatEndpoints(testCase)
-% The v1 distance_metadata is FLAT (ontologyNode_A/_B, integerIDs_A/_B,
-% ontologyStringValues_A/_B, ontologyNumericValues_A/_B empty by design, units).
-% The J migrator reshapes it into the nested `endpoints` array the V_eta schema
-% requires -- 1 -> 1. It does NOT emit a length_observation (no scalar distance in
-% the metadata doc; the value is the linked element's timeseries -- deferred Part
-% B) and does NOT mint an endpoint relation (the endpoint ids are pre-migration
-% ids that dangle single-doc -- Part A reverted). See
-% schemas/V_eta_distance_metadata_plan.md.
+function testDistanceMetadataIsCarriedWholeAsV1(testCase)
+% INVERTED 2026-10-29 from testDistanceMetadataReshapesFlatEndpoints, which
+% asserted that the FLAT v1 distance_metadata (ontologyNode_A/_B,
+% integerIDs_A/_B, ontologyStringValues_A/_B, ontologyNumericValues_A/_B,
+% units) was reshaped 1 -> 1 into the V_eta retire class `distance_metadata`
+% (a nested `endpoints` array + a `units` term). Since [reshaped retire
+% classes carried] (Jess Haley, 2026-10-29, did-schema
+% V_eta_go_forward_class_audit.md, "B"): v1 `distance_metadata` is "carried
+% whole as v1 for now, not reshaped into [its] retire class"; the reshape
+% waits for its V2 home. So the same document comes back as its untouched v1
+% source, noted under register row carry-147. What stays true: no
+% length_observation and no endpoint relation is written.
 v1 = struct();
 v1.document_class = struct('class_name', 'distance_metadata', 'class_version', '1.0.0', ...
     'superclasses', struct('class_name', 'base', 'class_version', '1.0.0'));
@@ -4106,28 +4109,25 @@ function testDistanceMetadataReshapesFlatEndpoints(testCase)
     'ontologyNumericValues_B', [],  'ontologyStringValues_B', 'uid_b1', ...
     'units', 'NCIT:C48367');
 
+bodies = did2.convert.migrators_j.distance_metadata(v1);
+verifyEqual(testCase, numel(bodies), 1, 'the v1 source only');
+verifyTrue(testCase, isequaln(bodies{1}, v1), 'returned untouched');
+
 out = runJ(v1);
-verifyEmpty(testCase, out.quarantine);
-verifyEqual(testCase, numel(out.migrated), 1);
+verifyCarriedWholeAsV1(testCase, v1, out);
 doc = out.migrated{1};
-verifyEqual(testCase, did2.unittest.helpers.classKey(doc), 'distance_metadata');
-
-blk = doc.get('distance_metadata');
-% two endpoint records, labelled A and B, in order
-verifyEqual(testCase, numel(blk.endpoints), 2);
-verifyEqual(testCase, blk.endpoints(1).label, 'A');
-verifyEqual(testCase, blk.endpoints(2).label, 'B');
-% the ontologyNode CURIE -> measurement.node; comma-split string ids -> array
-verifyEqual(testCase, blk.endpoints(1).measurement.node, 'animal_9');
-verifyEqual(testCase, blk.endpoints(2).measurement.node, 'patch_9');
-verifyEqual(testCase, numel(blk.endpoints(1).string_ids), 2);
-verifyEqual(testCase, blk.endpoints(2).integer_ids, [2 3]);
-% units CURIE -> ontology_term node
-verifyEqual(testCase, blk.units.node, 'NCIT:C48367');
-% the flat per-endpoint fields are gone (strict schema: only endpoints + units)
-verifyFalse(testCase, isfield(blk, 'ontologyNode_A'));
-verifyFalse(testCase, isfield(blk, 'ontology_node_a'));
-% NOT a length_observation and no endpoint relation minted (single-doc)
+verifyEqual(testCase, char(doc.get('document_class.schema_version')), 'did_v1');
+verifyEqual(testCase, char(doc.get('document_class.class_name')), 'distance_metadata');
+% the flat per-endpoint fields are kept as v1 wrote them; no endpoints array
+s = doc.toStruct();
+verifyEqual(testCase, s.distance_metadata.ontologyNode_A, 'animal_9');
+verifyEqual(testCase, s.distance_metadata.ontologyStringValues_B, 'uid_b1');
+verifyFalse(testCase, isfield(s.distance_metadata, 'endpoints'));
+verifyTrue(testCase, startsWith(out.summary.migration_notes.carried_as_v1.distinct_details{1}, ...
+    '[carry-147] awaiting_home: '), 'the note cites carry-147');
+verifyTrue(testCase, contains(out.summary.migration_notes.carried_as_v1.distinct_details{1}, ...
+    'reshaped retire classes carried'), 'the note names the decision');
+% NOT a length_observation and no endpoint relation minted
 names = cellfun(@(d) did2.unittest.helpers.classKey(d), out.migrated, 'UniformOutput', false);
 verifyFalse(testCase, any(strcmp(names, 'observation+length')));
 verifyFalse(testCase, any(strcmp(names, 'directed_relation')));
@@ -6217,152 +6217,104 @@ function testAnUnplacedSyncruleMappingIsCarriedUntouched(testCase)
 verifyTrue(testCase, isfield(out.summary.migration_notes, 'carried_as_v1'));
 end
 
-% ===================== oneepoch: a class that had NO schema ===============
+% ===================== oneepoch: carried whole as v1 =========================
 %
 % `oneepoch` reached 2026-08-10 with no V_eta schema, no migrator and no row on
-% any worklist, because coverage.py tagged it non-production -- which suppresses
-% the coverage gap. It is production: ndi.element.oneepoch concatenates an
-% element's N epochs into one, written at src/ndi/element.m:387 and read back at
-% +ndi/+element/oneepoch.m:78-80. A real document quarantined on "No schema file
-% for class oneepoch" (measured, scratch probe 8, run 31423494433).
+% any worklist, because coverage.py tagged it non-production. It is production:
+% ndi.element.oneepoch concatenates an element's N epochs into one, written at
+% src/ndi/element.m:387 and read back at +ndi/+element/oneepoch.m:78-80. Until
+% 2026-10-29 the migrator kept the class and folded its inherited
+% element_epoch block onto it (as `clocks`). Since [reshaped retire classes
+% carried] (Jess Haley, 2026-10-29, did-schema V_eta_go_forward_class_audit.md,
+% "B") every oneepoch is carried whole as v1: one on a global clock under
+% register row carry-148 (these tests), one stitched on device-local time
+% under carry-126 (testRawV1RoundTripOneEpoch).
 %
-% These drive the FULL pipeline via runJ, not the migrator directly. That matters:
-% the whole reason this migrator exists is that a BASE superclass migrator has
-% already rewritten the inherited block before the concrete migrator runs, and
-% calling migrators_j.oneepoch on a raw v1 body would test a shape that never
-% actually occurs.
-
-function testOneEpochFoldsItsInheritedBlockAndKeepsItsClass(testCase)
-% element_epoch is renamed to acquisition_epoch in V_eta, so no class of that
-% name exists for oneepoch to inherit from -- and the validator's top-level check
-% is strict. The inherited block must end up on the concrete one.
+% These drive the FULL pipeline via runJ, not the migrator directly.
+
+function v1 = oneEpochOnAGlobalClock(id)
+% a oneepoch whose clocks include a global one (utc)
 v1 = struct();
 v1.document_class = struct('class_name', 'oneepoch', 'class_version', '1.0.0', ...
     'superclasses', [ struct('class_name', 'element_epoch', 'class_version', '1.0.0'), ...
                       struct('class_name', 'base',          'class_version', '1.0.0'), ...
                       struct('class_name', 'epochid',       'class_version', '1.0.0')]);
 v1.depends_on = struct('name', {'element_id'}, 'value', {'elem_1'});
-v1.base = struct('id', 'oe_1', 'session_id', 'sess_09', ...
+v1.base = struct('id', id, 'session_id', 'sess_09', ...
     'name', '', 'datestamp', '2024-06-01T12:00:00.000Z');
 v1.epochid = struct('epochid', 'whole_session_ref1');
 v1.element_epoch = struct('epoch_clock', 'utc,dev_local_time', 't0_t1', [0 1; 2 3]);
 v1.oneepoch = struct('epoch_ids', 't00001,t00002,t00003');
+end
 
+function testOneEpochOnAGlobalClockIsCarriedWholeAsV1(testCase)
+% INVERTED 2026-10-29 from testOneEpochFoldsItsInheritedBlockAndKeepsItsClass,
+% which asserted the class was KEPT (`oneepoch`, stamped V2) with its
+% element_epoch block folded away. [reshaped retire classes carried] (Jess
+% Haley, 2026-10-29): it is carried whole as v1 until the global-clock
+% data-combination calculation is built, citing carry-148.
+v1 = oneEpochOnAGlobalClock('oe_1');
 out = runJ(v1);
-verifyEmpty(testCase, out.quarantine, 'oneepoch must not quarantine');
-verifyEqual(testCase, numel(out.migrated), 1);
+verifyCarriedWholeAsV1(testCase, v1, out);
 d = out.migrated{1};
-
-% the class is KEPT -- this is a fold, not a dissolution
-verifyEqual(testCase, did2.unittest.helpers.classKey(d), 'oneepoch');
-% the inherited block is GONE from the top level. Checked with isfield on the
-% struct rather than by expecting d.get() to throw -- probe 7's lesson: an
-% assertion about behaviour I have not observed is an assertion that can pass for
-% the wrong reason.
+verifyEqual(testCase, char(d.get('document_class.schema_version')), 'did_v1');
+verifyEqual(testCase, char(d.get('document_class.class_name')), 'oneepoch');
+% the inherited block is KEPT as v1 wrote it: nothing folded
 sOut = d.toStruct();
-verifyFalse(testCase, isfield(sOut, 'element_epoch'), ...
-    'the element_epoch block must not survive -- no V_eta class hosts it');
-verifyTrue(testCase, isfield(sOut, 'oneepoch'));
+verifyTrue(testCase, isfield(sOut, 'element_epoch'), 'the v1 element_epoch block is kept');
+verifyFalse(testCase, isfield(sOut.oneepoch, 'clocks'), 'nothing folded onto oneepoch');
+note = out.summary.migration_notes.carried_as_v1.distinct_details{1};
+verifyTrue(testCase, startsWith(note, '[carry-148] awaiting_home: '), note);
+verifyTrue(testCase, contains(note, 'reshaped retire classes carried'), note);
+end
+
+function testOneEpochCarryKeepsBothClocksAsV1WroteThem(testCase)
+% INVERTED 2026-10-29 from testOneEpochCarriesTheClocksTheBaseMigratorBuilt,
+% which asserted the base element_epoch migrator's `clocks` records landed on
+% `oneepoch.clocks`. The document is now carried whole as v1 ([reshaped retire
+% classes carried]): the multi-clock case (oneepoch.m:124 writes
+% strjoin(ecs,',')) comes back exactly as v1 wrote it, the comma-joined names
+% and the 2-by-N t0_t1.
+v1 = oneEpochOnAGlobalClock('oe_2');
+s = runJ(v1).migrated{1}.toStruct();
+verifyEqual(testCase, s.element_epoch.epoch_clock, 'utc,dev_local_time', ...
+    'both clocks in the comma-joined list must survive');
+verifyEqual(testCase, s.element_epoch.t0_t1, [0 1; 2 3]);
 end
 
-function testOneEpochCarriesTheClocksTheBaseMigratorBuilt(testCase)
-% The multi-clock case is REAL here: oneepoch.m:124 writes strjoin(ecs,','), every
-% clock the element had, where a plain element_epoch carries one. The base
-% element_epoch migrator parses the 2-by-N t0_t1 convention; this migrator must
-% MOVE that result without reshaping it.
-v1 = struct();
-v1.document_class = struct('class_name', 'oneepoch', 'class_version', '1.0.0', ...
-    'superclasses', [ struct('class_name', 'element_epoch', 'class_version', '1.0.0'), ...
-                      struct('class_name', 'base',          'class_version', '1.0.0'), ...
-                      struct('class_name', 'epochid',       'class_version', '1.0.0')]);
-v1.depends_on = struct('name', {'element_id'}, 'value', {'elem_1'});
-v1.base = struct('id', 'oe_2', 'session_id', 'sess_09', ...
-    'name', '', 'datestamp', '2024-06-01T12:00:00.000Z');
-v1.epochid = struct('epochid', 'whole_session_ref1');
-v1.element_epoch = struct('epoch_clock', 'utc,dev_local_time', 't0_t1', [0 1; 2 3]);
-v1.oneepoch = struct('epoch_ids', 't00001,t00002,t00003');
-
-d = runJ(v1).migrated{1};
-clocks = d.get('oneepoch.clocks');
-verifyEqual(testCase, numel(clocks), 2, ...
-    'both clocks in the comma-joined list must survive');
-verifyEqual(testCase, clocks(1).name, 'utc');
-verifyEqual(testCase, clocks(1).t0, 0);
-verifyEqual(testCase, clocks(1).t1, 2);
-verifyEqual(testCase, clocks(2).name, 'dev_local_time');
-verifyEqual(testCase, clocks(2).t0, 1);
-verifyEqual(testCase, clocks(2).t1, 3);
-end
-
-function testOneEpochKeepsTheSourceEpochIdsAndTheElementEdge(testCase)
-% epoch_ids is the ONLY field the class declares itself and the whole reason it is
-% distinct from element_epoch -- under fork A1 it becomes the derived_from_# edges.
-% Losing it would leave a concatenation that cannot say what it concatenated.
-v1 = struct();
-v1.document_class = struct('class_name', 'oneepoch', 'class_version', '1.0.0', ...
-    'superclasses', [ struct('class_name', 'element_epoch', 'class_version', '1.0.0'), ...
-                      struct('class_name', 'base',          'class_version', '1.0.0'), ...
-                      struct('class_name', 'epochid',       'class_version', '1.0.0')]);
-v1.depends_on = struct('name', {'element_id'}, 'value', {'elem_1'});
-v1.base = struct('id', 'oe_3', 'session_id', 'sess_09', ...
-    'name', '', 'datestamp', '2024-06-01T12:00:00.000Z');
-v1.epochid = struct('epochid', 'whole_session_ref1');
-% A GLOBAL clock beside the local one (2026-10-29): a oneepoch stitched on
-% device-local time alone is carried whole as v1 ([oneepoch method is data
-% combination], Jess Haley, did-schema V_eta_go_forward_class_audit.md;
-% testRawV1RoundTripOneEpoch), and this test is about the kept class.
+function testOneEpochCarryKeepsTheSourceEpochIdsAndTheElementEdge(testCase)
+% INVERTED 2026-10-29 from testOneEpochKeepsTheSourceEpochIdsAndTheElementEdge,
+% which asserted them on the kept class. epoch_ids is the ONLY field the class
+% declares itself; the calculation that replaces the carry turns it into its
+% per-epoch input_id links. On the carry it, the SYNTHETIC epoch id
+% (sourceCensus tracks that string as a grouping hazard, citing oneepoch.m:42)
+% and the element_id edge are all the v1 source's.
+v1 = oneEpochOnAGlobalClock('oe_3');
 v1.element_epoch = struct('epoch_clock', 'utc,dev_local_time', 't0_t1', [0 1; 930.35 931.35]);
-v1.oneepoch = struct('epoch_ids', 't00001,t00002,t00003');
-
-d = runJ(v1).migrated{1};
-verifyEqual(testCase, did2.unittest.helpers.classKey(d), 'oneepoch');
-verifyEqual(testCase, d.get('oneepoch.epoch_ids'), 't00001,t00002,t00003');
-% the SYNTHETIC epoch id survives too -- sourceCensus tracks exactly this string
-% as a grouping hazard, citing oneepoch.m:42, so it must not be quietly dropped
-verifyEqual(testCase, d.get('epochid.epochid'), 'whole_session_ref1');
-% READ THE EDGE TOLERANTLY. A depends_on entry is spelled `value` on a body a
-% migrator built and `document_id` once universalRenames has normalised it
-% (universalRenames.m:372-380), and BOTH shapes are live -- this document is a
-% passthrough, so its edge came through the rename and carries `document_id`
-% (observed: scratch probe 9, run 31424544834, printed `fieldnames: name,
-% document_id`).
-%
-% Reading one spelling is what broke this test twice: first via `depValue`, which
-% read only `value`, and then via `depVal`, which reads only `value` too -- I
-% swapped helpers on the INFERENCE that get() and toStruct() disagreed, without
-% ever observing get(). That inference was the error, not the first helper.
-%
-% The tolerant read is the codebase's own convention, not an invention here:
-% +did2/+validate/references.m:176-179 takes `document_id` when present and falls
-% back to `value`, with exactly this precedence. depValue now does the same.
-verifyEqual(testCase, depValue(d.toStruct(), 'element_id'), 'elem_1');
-end
-
-function testOneEpochValidatesAgainstItsNewTombstone(testCase)
-% THE POINT OF THE WHOLE CHANGE, asserted with validation ON rather than inferred
-% from the three tests above. Before the tombstone existed this quarantined with
-% "No schema file for class oneepoch"; the fold and the schema have to agree, and
-% only a validating run proves they do. runJ deliberately passes Validate=false,
-% so this calls v1_to_v2 directly.
-v1 = struct();
-v1.document_class = struct('class_name', 'oneepoch', 'class_version', '1.0.0', ...
-    'superclasses', [ struct('class_name', 'element_epoch', 'class_version', '1.0.0'), ...
-                      struct('class_name', 'base',          'class_version', '1.0.0'), ...
-                      struct('class_name', 'epochid',       'class_version', '1.0.0')]);
-v1.depends_on = struct('name', {'element_id'}, 'value', {'elem_1'});
-v1.base = struct('id', 'oe_4', 'session_id', 'sess_09', ...
-    'name', '', 'datestamp', '2024-06-01T12:00:00.000Z');
-v1.epochid = struct('epochid', 'whole_session_ref1');
-v1.element_epoch = struct('epoch_clock', 'utc,dev_local_time', 't0_t1', [0 1; 2 3]);
-v1.oneepoch = struct('epoch_ids', 't00001,t00002,t00003');
-
+out = runJ(v1);
+verifyCarriedWholeAsV1(testCase, v1, out);
+s = out.migrated{1}.toStruct();
+verifyEqual(testCase, s.oneepoch.epoch_ids, 't00001,t00002,t00003');
+verifyEqual(testCase, s.epochid.epochid, 'whole_session_ref1');
+% read tolerantly: depValue takes `document_id` when present, else `value`
+verifyEqual(testCase, depValue(s, 'element_id'), 'elem_1');
+end
+
+function testOneEpochOnAGlobalClockIsCarriedUnderValidation(testCase)
+% INVERTED 2026-10-29 from testOneEpochValidatesAgainstItsNewTombstone, which
+% asserted the kept class validated against its V_eta tombstone. A carried
+% document is v1 and is never validated against V_eta (v1_to_v2: "A 'did_v1'
+% body is never validated against V_eta"), so with validation ON it must still
+% not quarantine, and comes back carried.
+v1 = oneEpochOnAGlobalClock('oe_4');
 out = did2.convert.v1_to_v2(v1, 'Validate', true, 'TargetVersion', 'V_eta');
 if ~isempty(out.quarantine)
     verifyFail(testCase, sprintf('oneepoch quarantined under validation: %s', ...
         out.quarantine(1).reason));
 end
 verifyEqual(testCase, numel(out.migrated), 1);
-verifyEqual(testCase, did2.unittest.helpers.classKey(out.migrated{1}), 'oneepoch');
+verifyEqual(testCase, char(out.migrated{1}.get('document_class.schema_version')), 'did_v1');
+verifyEqual(testCase, char(out.migrated{1}.get('document_class.class_name')), 'oneepoch');
 end
 
 function v = depValue(b, name)
diff --git a/tests/+did2/+unittest/testRawV1RoundTripOneEpoch.m b/tests/+did2/+unittest/testRawV1RoundTripOneEpoch.m
index eba1248f..8e23052e 100644
--- a/tests/+did2/+unittest/testRawV1RoundTripOneEpoch.m
+++ b/tests/+did2/+unittest/testRawV1RoundTripOneEpoch.m
@@ -16,13 +16,14 @@
 %   and comma-joined clock names, as oneepoch.m writes them). No corpus of the
 %   four zips holds one.
 %
-%   The migrator keeps a oneepoch on a global clock (class oneepoch), its
-%   element_epoch block folded onto oneepoch as `clocks` by
-%   did2.convert.migrators.element_epoch, which the raw migrator now calls
-%   itself. No known findings. A oneepoch stitched on device-local time (no
-%   clock of it global) is carried whole as v1, needs_mapping ([oneepoch is a
-%   redundant calculation], Jess Haley, 2026-10-27; [oneepoch method is data
+%   EVERY oneepoch is carried whole as v1. One stitched on device-local time
+%   (no clock of it global) is carried needs_mapping, carry-126 ([oneepoch is
+%   a redundant calculation], Jess Haley, 2026-10-27; [oneepoch method is data
 %   combination], Jess Haley, 2026-10-29): testADeviceLocalOneEpochIsCarried.
+%   One on a global clock, which until 2026-10-29 kept the class `oneepoch`
+%   with its element_epoch block folded onto it as `clocks`, is carried
+%   awaiting_home, carry-148 ([reshaped retire classes carried], Jess Haley,
+%   2026-10-29, "B"): testAGlobalClockOneEpochIsCarried.
 tests = functiontests(localfunctions);
 end
 
@@ -35,7 +36,10 @@ function testWriterShapesRoundTrip(testCase)
 % was converted; it is a oneepoch stitched on device-local time, now carried
 % whole as v1 ([oneepoch is a redundant calculation], Jess Haley, 2026-10-27;
 % [oneepoch method is data combination], Jess Haley, 2026-10-29, did-schema
-% V_eta_go_forward_class_audit.md). CARRIED below says which.
+% V_eta_go_forward_class_audit.md). INVERTED 2026-10-29 in full: the
+% two-clock fixture (one of them utc) kept the class; it is carried too
+% ([reshaped retire classes carried], Jess Haley, 2026-10-29). CARRIED below
+% says which.
 f = fixtures();
 for k = 1:size(f, 1)
     [label, v1, carried] = f{k, :};
@@ -61,25 +65,42 @@ function testADeviceLocalOneEpochIsCarried(testCase)
 verifyTrue(testCase, contains(notes, 'needs_mapping'), notes);
 verifyTrue(testCase, startsWith(notes, '[carry-126] needs_mapping: '), notes);
 verifyTrue(testCase, contains(notes, 'device-local time'), notes);
-% any global clock among its clocks: stitched on that clock, not carried
+end
+
+function testAGlobalClockOneEpochIsCarried(testCase)
+% INVERTED 2026-10-29 from the second half of testADeviceLocalOneEpochIsCarried,
+% which asserted that a oneepoch with any global clock among its clocks was
+% NOT carried (it kept the class `oneepoch`). [reshaped retire classes
+% carried] (Jess Haley, 2026-10-29, did-schema V_eta_go_forward_class_audit.md,
+% "B"): it is carried whole as v1, awaiting_home, citing register row
+% carry-148, until the global-clock data-combination calculation is built.
+% It is still told apart from the device-local carry: its note is not
+% carry-126's.
 cases = { ...
     'dev_local_time,utc', [0 1.7e9; 100 1.7e9 + 100]; ...
     'dev_local_time,approx_dev_global_time', [0 7.3855e5; 100 7.3855e5 + 0.01]; ...
     'exp_global_time', [7.3855e5; 7.3855e5 + 0.01]};
 for k = 1:size(cases, 1)
     [clock, t] = cases{k, :};
-    out = runRaw(oneEpoch(clock, t));
-    verifyEmpty(testCase, out.passthrough, [clock ': not carried']);
-    assertNotEmpty(testCase, out.migrated, clock);
-    verifyEqual(testCase, out.migrated{1}.className(), 'oneepoch', clock);
+    [carried, notes, out] = did2.unittest.helpers.carryNote({oneEpoch(clock, t)});
+    verifyTrue(testCase, carried, [clock ': carried as the untouched v1 source']);
+    verifyEqual(testCase, numel(out.passthrough), 1, [clock ': recorded as a passthrough']);
+    verifyTrue(testCase, startsWith(notes, '[carry-148] awaiting_home: '), [clock ': ' notes]);
+    verifyTrue(testCase, contains(notes, 'reshaped retire classes carried'), [clock ': ' notes]);
+    verifyFalse(testCase, contains(notes, 'carry-126'), [clock ': ' notes]);
 end
 end
 
 function testADifferentSourceIsReportedLost(testCase)
 % (a global-clock oneepoch since 2026-10-29: the device-local one is carried,
-% [oneepoch method is data combination])
+% [oneepoch method is data combination]). INVERTED 2026-10-29: the global-clock
+% one is carried too ([reshaped retire classes carried], Jess Haley,
+% 2026-10-29), so the round trip reads the carry, and a source that differs
+% from the carried body is still reported lost at the field that differs.
 v1 = oneEpoch('dev_local_time,utc', [0 1.7e9; 100 1.7e9 + 100]);
 out = runRaw(v1);
+r = did2.convert.roundtrip.check(v1, out.migrated);
+verifyTrue(testCase, r.carried, 'the global-clock oneepoch is carried');
 v1.oneepoch.epoch_ids = 't00001';
 r = did2.convert.roundtrip.check(v1, out.migrated);
 verifyEqual(testCase, {r.lost.path}, {'oneepoch.epoch_ids'});
@@ -89,10 +110,11 @@ function testADifferentSourceIsReportedLost(testCase)
 
 function f = fixtures()
 % {label, v1, carried}: a oneepoch with no global clock is carried
-% ([oneepoch method is data combination], Jess Haley, 2026-10-29)
+% ([oneepoch method is data combination], Jess Haley, 2026-10-29); so is one
+% on a global clock ([reshaped retire classes carried], Jess Haley, 2026-10-29)
 f = { ...
     'oneepoch, one clock', oneEpoch('dev_local_time', [0; 100]), true; ...
-    'oneepoch, two clocks', oneEpoch('dev_local_time,utc', [0 1.7e9; 100 1.7e9 + 100]), false};
+    'oneepoch, two clocks', oneEpoch('dev_local_time,utc', [0 1.7e9; 100 1.7e9 + 100]), true};
 end
 
 function d = oneEpoch(clock, t0t1)
diff --git a/tests/+did2/+unittest/testRawV1RoundTripPosition.m b/tests/+did2/+unittest/testRawV1RoundTripPosition.m
index b6305459..39a2e687 100644
--- a/tests/+did2/+unittest/testRawV1RoundTripPosition.m
+++ b/tests/+did2/+unittest/testRawV1RoundTripPosition.m
@@ -39,9 +39,12 @@
 %                        is written"; the metadata and its element's values
 %                        become one position calculation, and until position
 %                        elements are mapped (carry-063) both stay carried.
-%     distance_metadata  converts to the V2 distance_metadata (a class whose
-%                        disposition is `retire`; a question in the phase
-%                        report) and loses nothing.
+%     distance_metadata  carried (carry-147): until 2026-10-29 it converted
+%                        to the V2 distance_metadata (a class whose
+%                        disposition is `retire`); since [reshaped retire
+%                        classes carried] (Jess Haley, 2026-10-29, did-schema
+%                        V_eta_go_forward_class_audit.md, "B") it is carried
+%                        whole as v1 until its V2 home is decided.
 %     both, beside their element   the element's type (`position`,
 %                        `distance`, on ndi.element.timeseries) is not on the
 %                        element type list, so it is carried (carry-063) and
@@ -58,8 +61,8 @@ function testWriterDocumentsRoundTrip(testCase)
     'position writer output', {position('pm_1', 'EMPTY:0000137')}, true; ...
     'position, converted shape', {positionConvertedShape()}, true; ...  % carried since [position and its metadata are one statement]
     'position beside its carried element', {position('pm_1', 'EMPTY:0000137'), track('el_pos', 'position')}, true; ...
-    'distance writer output (19 patches)', {distance('dm_1', 1:19)}, false; ...
-    'distance writer output (one patch)', {distance('dm_2', 4)}, false; ...
+    'distance writer output (19 patches)', {distance('dm_1', 1:19)}, true; ...  % carried since [reshaped retire classes carried]
+    'distance writer output (one patch)', {distance('dm_2', 4)}, true; ...
     'distance beside its carried element', {distance('dm_1', 1:19), track('el_dist', 'distance')}, true};
 for k = 1:size(cases, 1)
     [label, batch, carried] = cases{k, :};
@@ -76,6 +79,8 @@ function testWriterDocumentsRoundTrip(testCase)
 end
 
 function testEveryWriterFieldOfADistanceIsAccountedFor(testCase)
+% (since [reshaped retire classes carried], Jess Haley, 2026-10-29, the
+% document is carried whole as v1, so every field is matched on the carry)
 v1 = distance('dm_1', 1:19);
 r = did2.convert.roundtrip.check(v1, runRaw({v1}).migrated);
 accounted = [r.matched, r.restated, r.equivalent, r.not_set, r.exempt];
@@ -93,8 +98,16 @@ function testCarriesCiteTheirRows(testCase)
     '[carry-053] unplaced: position_metadata.units, position_metadata.dimensions');
 verifyCarriedFor(testCase, {position('pm_1', 'EMPTY:0000137'), track('el_pos', 'position')}, ...
     '[carry-053]');
+% INVERTED 2026-10-29: beside its carried element the distance_metadata cited
+% carry-064 ([dependents of a carried document carry]); since [reshaped
+% retire classes carried] (Jess Haley, 2026-10-29) its own migrator carries it
+% first, under carry-147, and v1_to_v2's carryDependents follows only a
+% document that was converted
+verifyCarriedFor(testCase, {distance('dm_1', 1:19)}, ...
+    '[carry-147] awaiting_home: distance_metadata');
+verifyCarriedFor(testCase, {distance('dm_2', 4)}, 'reshaped retire classes carried');
 verifyCarriedFor(testCase, {distance('dm_1', 1:19), track('el_dist', 'distance')}, ...
-    '[carry-064] needs_mapping: its element_id edge names el_dist');
+    '[carry-147] awaiting_home: distance_metadata');
 end
 
 % ===================== the switch ======================================
@@ -149,15 +162,19 @@ function testAPositionNamingNoMeasurementIsCarried(testCase)
 verifyTrue(testCase, isCarried(old, []), 'not carried on the renamed path');
 end
 
-function testTheConvertedDocumentsValidate(testCase)
-% positionConvertedShape() is no longer a converted document ([position and
-% its metadata are one statement], Jess Haley, 2026-10-27): it is carried, and
-% its carry is checked by testAPositionWithOnlyABodyPartIsCarried
+function testTheCarriedDistancesPassValidation(testCase)
+% INVERTED 2026-10-29 from testTheConvertedDocumentsValidate, which validated
+% the V2 distance_metadata the migrator wrote. Since [reshaped retire classes
+% carried] (Jess Haley, 2026-10-29, did-schema V_eta_go_forward_class_audit.md)
+% no distance_metadata is converted: with validation ON the writer's documents
+% must still not quarantine, and come back as their untouched v1 sources.
+% (positionConvertedShape() is carried too, [position and its metadata are one
+% statement]; its carry is checked by testAPositionWithOnlyABodyPartIsCarried.)
 for b = {{distance('dm_1', 1:19)}, {distance('dm_2', 4)}}
     out = did2.convert.v1_to_v2(b{1}, 'Validate', true, 'TargetVersion', 'V_eta');
-    label = b{1}{1}.document_class.class_name;
+    label = b{1}{1}.base.id;
     verifyEmpty(testCase, out.quarantine, sprintf('%s: %s', label, quarantineText(out)));
-    verifyNotEmpty(testCase, out.migrated, label);
+    verifyTrue(testCase, isCarried(out, b{1}{1}), [label ': not carried as the untouched v1 source']);
 end
 end
 
@@ -187,18 +204,17 @@ function testAFieldTheMigratorDoesNotReadCarriesTheDocument(testCase)
 % ===================== a check that cannot fail proves nothing ===========
 
 function testABrokenEmissionIsReportedLost(testCase)
-% distance: a patch number and an id changed
+% distance: a patch number and an id changed. INVERTED 2026-10-29: it broke
+% the V2 `endpoints` the migrator wrote; since [reshaped retire classes
+% carried] (Jess Haley, 2026-10-29) the document is carried whole as v1, so
+% the same two values are changed on the carried body
 v1 = distance('dm_1', 1:19);
-docs = structs(runRaw({v1}).migrated);
-ep = docs{1}.distance_metadata.endpoints;
-if iscell(ep)
-    ep{2}.integer_ids(3) = 99;
-    ep{1}.measurement.node = 'EMPTY:0000097';
-else
-    ep(2).integer_ids(3) = 99;
-    ep(1).measurement.node = 'EMPTY:0000097';
-end
-docs{1}.distance_metadata.endpoints = ep;
+out = runRaw({v1});
+docs = structs(out.migrated);
+assertEqual(testCase, numel(docs), 1, 'the v1 source only');
+verifyTrue(testCase, did2.convert.roundtrip.check(v1, docs).carried);
+docs{1}.distance_metadata.integerIDs_B(3) = 99;
+docs{1}.distance_metadata.ontologyNode_A = 'EMPTY:0000097';
 r = did2.convert.roundtrip.check(v1, docs);
 verifyEqual(testCase, sort(lostPaths(r)), ...
     sort({'distance_metadata.integerIDs_B', 'distance_metadata.ontologyNode_A'}));
diff --git a/tests/+did2/+unittest/testResolveEpochProbemap.m b/tests/+did2/+unittest/testResolveEpochProbemap.m
index 22ef46fb..3dc22c76 100644
--- a/tests/+did2/+unittest/testResolveEpochProbemap.m
+++ b/tests/+did2/+unittest/testResolveEpochProbemap.m
@@ -552,7 +552,8 @@ function testJHPositionAndDistanceCarryWithTheirEpochs(testCase)
 % [element types have no default] (Jess Haley, 2026-10-26): `position` and
 % `distance` are on no element type list, so the elements are carried as v1
 % (carry-063), and [dependents of a carried document carry] carries their
-% element_epochs and the distance_metadata (carry-064). No acquisition_epoch,
+% element_epochs (carry-064); the distance_metadata is carried by its own
+% migrator (carry-147, [reshaped retire classes carried]). No acquisition_epoch,
 % track calculation, coordinate system or nearest-patch item is written.
 sid = 'sess_jh';
 worm = subjectBody('worm_1', sid, 'N2_worm0001');