docs: drop project-phase and ticket-id references from this PR's code comments

Comments shouldn't carry 'Stage A/B/C/D' or 'DRC-XXXX' — they're write-time
context that means nothing to a reader later. Reworded the comments and
docstrings this PR adds to describe the behavior in plain terms instead.
Ticket/phase references stay in commit messages and Linear, not the code.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Signed-off-by: Danyel Fisher <danyel@gmail.com>

Author: danyelf

js/packages/storybook/stories/profile-distribution/InlineProfileDistributionCellDatetime.stories.tsx

@@ -5,7 +5,7 @@ import { continuousEventTime, continuousStable } from "./fixtures";
 import { SchemaContainerMock, SchemaRowMock } from "./SchemaRowMock";
 
 /**
- * Continuous tooltip formatting by column type (DRC-3390 review note 1).
+ * Continuous tooltip formatting by column type.
  *
  * The backend wraps datetime columns in an `epoch()` cast before binning, but
  * the numeric edges mean different things per type:

js/packages/storybook/stories/profile-distribution/fixtures.ts

@@ -28,11 +28,11 @@
 // The top-K (counts) and result types below are deliberately Storybook-LOCAL
 // and DIVERGE from the canonical `run.ts` contract: the fixtures bundle
 // envelope-level `base_total`/`current_total` onto the counts payload for
-// Stage-A display convenience, whereas the wire contract puts those on the
+// display convenience here, whereas the wire contract puts those on the
 // `ProfileDistributionOkResult` envelope and types counts as `(number|null)[]`.
 // Do NOT copy these shapes into production — use the `@datarecce/ui/api` types.
 // (Fully canonicalizing the counts fixtures means moving totals to the envelope
-// across the Stage-A discrete stories; deferred — storybook-only, no prod impact.)
+// across the discrete stories; deferred — storybook-only, no prod impact.)
 
 import type { ProfileDistributionHistogramPayload } from "@datarecce/ui/api";
 
@@ -183,7 +183,7 @@ export const continuousAddedOnly = continuousFromProportions(
  * backend's `epoch()` cast emits for `TIME` (distinct from the Unix-epoch
  * seconds a TIMESTAMP/DATE emits). An activity bump around midday. Used to
  * exercise the `HH:MM:SS` clock-time tooltip path (vs the calendar-date path
- * a TIMESTAMP takes) — see DRC-3390 review note 1.
+ * a TIMESTAMP takes).
  */
 const HOUR = 3600;
 const timeOfDayEdges = [0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22].map(

js/packages/ui/src/components/data/InlineProfileDistributionCell.tsx

@@ -19,10 +19,10 @@ import {
 
 /**
  * @file InlineProfileDistributionCell.tsx
- * @description DRC-3390 Stage C — the schema-grid cell container that turns a
- * single column's {@link ProfileDistributionColumnPayload} into the right
- * Stage A paired-histogram leaf, and renders the loading / error / empty /
- * per-column-failure states around it.
+ * @description The schema-grid cell container that turns a single column's
+ * {@link ProfileDistributionColumnPayload} into the right paired-histogram
+ * leaf, and renders the loading / error / empty / per-column-failure states
+ * around it.
  *
  * This is the one place the snake_case wire payload is mapped to the
  * camelCase cell props, and the only place that has to know the column's dbt
@@ -50,10 +50,10 @@ export interface InlineProfileDistributionCellProps {
   columnType?: string;
   /**
    * Envelope-level row totals — the denominator for counts-mode proportions.
-   * Unused in ranks mode (the DuckDB Stage B path that ships today), so
-   * optional. NOT speculative: Stage D's full adapters (Snowflake/BigQuery
-   * top-K return value+count pairs) emit counts mode, so this is threaded now
-   * to match the single `run.ts` payload contract rather than retrofitted then.
+   * Unused in ranks mode (the DuckDB path that ships today), so optional. NOT
+   * speculative: full warehouse adapters (Snowflake/BigQuery top-K return
+   * value+count pairs) emit counts mode, so this is threaded now to match the
+   * single `run.ts` payload contract rather than retrofitted later.
    */
   baseTotal?: number;
   currentTotal?: number;
@@ -65,9 +65,9 @@ export interface InlineProfileDistributionCellProps {
 }
 
 /**
- * Calendar-date types whose histogram edges are seconds since the Unix epoch
- * (see DRC-3504). Bare `TIME` is deliberately excluded — its edges are
- * seconds-since-midnight, not an epoch, and is handled by `isTimeOfDayType`.
+ * Calendar-date types whose histogram edges are seconds since the Unix epoch.
+ * Bare `TIME` is deliberately excluded — its edges are seconds-since-midnight,
+ * not an epoch, and is handled by `isTimeOfDayType`.
  */
 function isDatetimeType(type?: string): boolean {
   if (!type) return false;
@@ -81,8 +81,8 @@ function isDatetimeType(type?: string): boolean {
  * Time-of-day types (`TIME`, `TIME WITH/WITHOUT TIME ZONE`). The backend's
  * `epoch()` cast emits **seconds-since-midnight** (0–86399) for these, so the
  * edges must be read as a clock time, not a calendar date — otherwise every
- * tooltip collapses to "Jan 1, 1970" (DRC-3390 review note 1). Matches `time`
- * but not `timestamp`/`datetime`, which carry real epoch seconds.
+ * tooltip collapses to "Jan 1, 1970". Matches `time` but not
+ * `timestamp`/`datetime`, which carry real epoch seconds.
  */
 function isTimeOfDayType(type?: string): boolean {
   if (!type) return false;
@@ -152,12 +152,12 @@ function toDiscreteData(
       trimmed: p.trimmed,
     };
   }
-  // Counts mode. Stage B (DuckDB) only emits ranks, so this branch is inert
-  // today — but it's a known future need, not speculation: Stage D's full
+  // Counts mode. The DuckDB path only emits ranks, so this branch is inert
+  // today — but it's a known future need, not speculation: full warehouse
   // adapters (Snowflake/BigQuery top-K return value+count pairs) emit counts,
-  // and the cell conforms to the whole `run.ts` payload contract now so Stage D
-  // is a backend-only change. A per-slot `null` means the value is absent from
-  // that env's top-K — coerce to 0 so the bar simply doesn't render
+  // and the cell conforms to the whole `run.ts` payload contract now so adding
+  // them is a backend-only change. A per-slot `null` means the value is absent
+  // from that env's top-K — coerce to 0 so the bar simply doesn't render
   // (gap-on-absent), which is exactly the cell's 0-height behavior.
   return {
     mode: "counts",

js/packages/ui/src/components/data/__tests__/InlineProfileDistributionCell.test.tsx

@@ -1,8 +1,8 @@
 /**
  * @file InlineProfileDistributionCell.test.tsx
- * @description Tests for the Stage C schema-grid cell container (DRC-3390).
- * Covers the state machine: loading / error / empty / per-column failure /
- * histogram / topk (ranks + counts).
+ * @description Tests for the schema-grid cell container. Covers the state
+ * machine: loading / error / empty / per-column failure / histogram /
+ * topk (ranks + counts).
  */
 
 import { render } from "@testing-library/react";
@@ -146,7 +146,7 @@ describe("InlineProfileDistributionCell", () => {
   it("formats TIME histogram edges as HH:MM:SS clock times, not dates", () => {
     // The backend's epoch() cast emits seconds-since-midnight for TIME, so the
     // tooltip must read as a clock time — never the bogus "Jan 1, 1970" a
-    // calendar-date formatter would produce (DRC-3390 review note 1).
+    // calendar-date formatter would produce.
     const timeHistogram: ProfileDistributionHistogramPayload = {
       ...histogram,
       // 00:00:00, 01:00:00, 02:00:00, ... (whole hours past midnight)

js/packages/ui/src/components/schema/SchemaView.tsx

@@ -240,8 +240,8 @@ export function PrivateSchemaView(
     new Map(),
   );
 
-  // DRC-3390: per-model opt-in to profile *every* column rather than just the
-  // changed ones. Keyed by the node id it was enabled for (not a bare boolean +
+  // Per-model opt-in to profile *every* column rather than just the changed
+  // ones. Keyed by the node id it was enabled for (not a bare boolean +
   // reset-in-effect) so switching nodes restores the changed-columns default on
   // the SAME render — the view isn't remounted per node, so a reset effect left
   // one stale render that fired a spurious all-columns run before it landed.

js/packages/ui/src/components/schema/selectInlineProfileScope.ts

@@ -1,8 +1,8 @@
 /**
  * @file selectInlineProfileScope.ts
- * @description DRC-3390 Stage C — pure scoping logic for the inline paired
- * distribution in the schema view, extracted from `SchemaView` so the wiring
- * is unit-testable in isolation (no React, no lineage context).
+ * @description Pure scoping logic for the inline paired distribution in the
+ * schema view, extracted from `SchemaView` so the wiring is unit-testable in
+ * isolation (no React, no lineage context).
  *
  * Given the change signals for a node and the user's "Profile all columns"
  * opt-in, it decides which columns to profile and whether to profile at all.
@@ -90,7 +90,7 @@ export function selectInlineProfileScope({
   // Attribute impacted ids to this node by exact `<nodeId>_<column>` membership
   // over the node's own columns (shared with the grid's isImpacted painting) —
   // never by prefix-stripping the global set, which would mis-attribute a
-  // sibling model's columns (DRC-3390 review #1; identity cleanup: DRC-3646).
+  // sibling model's columns.
   if (impactedColumns && nodeId && nodeColumnNames) {
     for (const name of nodeColumnNames) {
       if (isColumnImpacted(nodeId, name, impactedColumns)) names.add(name);
@@ -105,11 +105,11 @@ export function selectInlineProfileScope({
   // specific columns, so any column's *values* may have shifted even when its
   // definition is untouched, and the changed-column subset would under-cover.
   //
-  // COST NOTE (intended for Stage D — DRC-3631): combined with the run firing
-  // on node-open, this means opening a whole-model-changed node profiles ALL
-  // columns with no count cap. Cheap on the DuckDB-only Stage B path; against a
-  // wide warehouse model this is an unbounded scan triggered by navigation. An
-  // adapter-cost guard / column cap is deferred to the Stage D adapter work.
+  // COST NOTE: combined with the run firing on node-open, opening a
+  // whole-model-changed node profiles ALL columns with no count cap. Cheap on
+  // the in-process DuckDB path; against a wide warehouse model this is an
+  // unbounded scan triggered by navigation. A column cap / adapter-cost guard
+  // is deferred to the warehouse-adapter work (tracked separately).
   const scopedColumns =
     profileAllColumns || wholeModelChange || !hasChangedScope
       ? undefined

js/packages/ui/src/hooks/useInlineProfileDistribution.ts

@@ -13,7 +13,7 @@ import { useApiConfig } from "./useApiConfig";
 
 /**
  * @file useInlineProfileDistribution.ts
- * @description DRC-3390 Stage C — fire-and-parse hook for inline paired
+ * @description Fire-and-parse hook for inline paired
  * distributions in the schema view.
  *
  * Gated end-to-end on the `inline_profile` server flag: when the flag is off
@@ -29,7 +29,7 @@ import { useApiConfig } from "./useApiConfig";
  * hand-rolled submit/poll loop. CAVEAT: the OSS fetch client sets no request
  * timeout (`useApiConfig.ts`), so a slow remote-warehouse adapter would hold
  * this request open with no client-side abort — there is NO timeout error or
- * Sentry capture for that case today (tracked in DRC-3629). Transport errors
+ * Sentry capture for that case today (tracked separately). Transport errors
  * that do occur (network failure, non-2xx) surface as the hook's `error` state
  * and a Sentry capture via the queryFn's catch.
  *

js/packages/ui/src/utils/__tests__/cllColumnId.test.ts

@@ -1,8 +1,7 @@
 /**
  * @file cllColumnId.test.ts
- * @description The CLL column-id grammar + safe membership (DRC-3390 review #1,
- * DRC-3646). Shared by the schema grid's isImpacted painting and the inline
- * profile-distribution scope.
+ * @description The CLL column-id grammar + safe membership, shared by the
+ * schema grid's isImpacted painting and the inline profile-distribution scope.
  */
 
 import { describe, expect, it } from "vitest";

js/packages/ui/src/utils/__tests__/formatTime.test.ts

@@ -1,9 +1,9 @@
 /**
  * @file formatTime.test.ts
  * @description Covers `formatTimeOfDay` — the wall-clock formatter for
- * seconds-since-midnight (DRC-3390 review note 1). Its contract differs from
- * `formatDuration("compact")`: clock times always show two-digit hours, never
- * drop the hours component, and wrap at 24h.
+ * seconds-since-midnight. Its contract differs from `formatDuration("compact")`:
+ * clock times always show two-digit hours, never drop the hours component, and
+ * wrap at 24h.
  */
 
 import { describe, expect, it } from "vitest";

js/packages/ui/src/utils/cllColumnId.ts

@@ -5,10 +5,11 @@
  * Column-level-lineage ids are `{node_id}_{column_name}` (mirrors the backend's
  * `recce/util/lineage.py`). Because both halves can contain underscores and one
  * node id can be a string-prefix of another, this id is NOT safely parseable by
- * splitting/prefix-stripping — that mis-attributes a sibling model's columns
- * (DRC-3390 review #1). Always go the other way: build the id from a KNOWN
- * (nodeId, column) pair and test membership. A fully-typed Node/Column identity
- * is tracked in DRC-3646; until then, route the join + membership through here.
+ * splitting/prefix-stripping — doing so mis-attributes a sibling model's columns
+ * (e.g. `model.shop.orders` absorbing a column of `model.shop.orders_summary`).
+ * Always go the other way: build the id from a KNOWN (nodeId, column) pair and
+ * test membership. A fully-typed Node/Column identity would remove the hazard
+ * entirely (tracked separately); until then, route the join + membership here.
  */
 
 /** Construct a CLL column id from a node's `unique_id` and a column name. */
@@ -27,7 +28,7 @@ export function cllColumnId(nodeId: string, column: string): string {
  * for a column the node doesn't have can collide with a sibling's. Callers
  * iterate the node's own columns (the schema grid's merged rows, the profile
  * scope's `nodeColumnNames`), so this never arises in practice. The lossless
- * fix is typed identity (DRC-3646).
+ * fix is typed identity.
  */
 export function isColumnImpacted(
   nodeId: string,