@rangojs/router 0.0.0-experimental.15 → 0.0.0-experimental.151

package/src/rsc/shell-capture.ts

@@ -0,0 +1,1716 @@
+/**
+ * PPR shell capture orchestration (Axis 2, see docs/design/ppr-shell-resume.md).
+ *
+ * Capture does NOT flow through the HTTP middleware pipeline. The integrated PPR
+ * serve path (rsc-rendering.ts + shell-serve.ts) builds a ShellCaptureDescriptor
+ * from the route's `ppr` path option after the served response is built and calls
+ * scheduleShellCapture. The capture then runs as a background task that re-derives
+ * the page via `ctx.router.match()` under its OWN derived request context — fresh
+ * handle store, `_shellCaptureRun: true` so loaders mask (loader-mask.ts) and every
+ * loading() subtree postpones. The render is MIXED-CHAIN: cache()'d segments replay
+ * from ring 3, uncached segments execute their handlers fresh. It drives the static
+ * prerender to a quiescent shell, aborts to freeze the prelude + postponed state,
+ * and stores the pair via putShell. Because it uses match() rather than the HTTP
+ * pipeline, the middleware chain (auth, logging) never re-runs — it already ran for
+ * the triggering request, and the derived context inherits its post-middleware
+ * state (variables, cache store). Guarding is serve-time.
+ */
+
+import React from "react";
+import { bufferToBase64 } from "../cache/cf/cf-base64.js";
+import { reportCacheError } from "../cache/cache-error.js";
+import { runBackground } from "../cache/background-task.js";
+import { enqueueSerializedCapture } from "./capture-queue.js";
+import { SHELL_CAPTURE_MAX_WAIT_MS } from "./shell-capture-constants.js";
+import { INTERNAL_RANGO_DEBUG } from "../internal-debug.js";
+import { observePhase, PHASES } from "../router/instrument.js";
+import {
+  runWithRequestContext,
+  setRequestContextParams,
+  wireRenderBarrier,
+  UNTRACKED_BACKGROUND_TASK,
+  type RequestContext,
+} from "../server/request-context.js";
+import { createHandleStore, type HandleStore } from "../server/handle-store.js";
+import {
+  maskNestedContainerThenables,
+  type MaskReport,
+} from "../router/segment-resolution/mask-nested.js";
+import { isInsideLoaderScope } from "../server/context.js";
+import { isThenable } from "../handles/is-thenable.js";
+import type {
+  ShellCacheEntry,
+  SegmentCacheStore,
+  ShellSnapshotRecord,
+} from "../cache/types.js";
+import {
+  elideLoaderContainer,
+  isLoaderHoleMarker,
+} from "../router/segment-resolution/loader-snapshot.js";
+import {
+  RecordingShellStore,
+  SnapshotOnlySegmentStore,
+  getRecordingStore,
+} from "../cache/shell-snapshot.js";
+import type { HandlerContext } from "./handler-context.js";
+import type { SSRModule } from "./types.js";
+import { buildFullPayload } from "./full-payload.js";
+import { resolveDeferredHandleValues } from "../handles/deferred-resolution.js";
+import { RSC_FLIGHT_ONLY_PHASES, renderRscFlightStage } from "./helpers.js";
+
+/**
+ * Task-quantized quiesce: the number of consecutive macrotask hops with zero new
+ * Flight bytes that marks the shell "quiet". This replaces the old 50ms
+ * wall-clock debounce.
+ *
+ * The capture Flight render is a REGULAR renderToReadableStream (not a static
+ * prerender), so React schedules both its retries and its byte-flush on
+ * setTimeout(0) MACROTASKS (verified against the vendored edge production
+ * react-server-dom build: pingTask uses scheduleMicrotask only when
+ * request.type === PRERENDER, otherwise setTimeout; enqueueFlush is always
+ * setTimeout). Masked loaders are the live lane — their rows never emit — so once
+ * the shell rows finish flushing the stream goes permanently byte-silent, and K
+ * consecutive quiet macrotask hops after the last observed byte declare quiesce.
+ *
+ * K=2 gives a race window of ~two event-loop turns: shell work still producing
+ * bytes keeps resetting the counter; anything not producing bytes within the
+ * window (the masked loaders, and any genuinely pending I/O) becomes a hole. The
+ * only residual is raw per-request I/O rendered directly in shell (not via a
+ * loader) that resolves inside the window — a documented shell anti-pattern; put
+ * per-request data in loaders. See docs/design/ppr-shell-resume.md.
+ */
+const FLIGHT_QUIET_HOPS = 2;
+
+/**
+ * Default capture budget. Canonical value, raise rationale, and ceiling math
+ * live on the leaf module (shell-capture-constants.ts, importable from the ssr
+ * graph too). Re-exported here so shell-build-manifest.ts's envelope math
+ * keeps its existing import site and cannot drift from the capture.
+ */
+export { SHELL_CAPTURE_MAX_WAIT_MS };
+
+/**
+ * Upper bound on waiting for the capture's DEFERRED cache writes to settle before
+ * draining the snapshot. Cache writes run under waitUntil (fire-and-forget on
+ * Node, executionContext on workerd), so a MISS-at-capture value's setItem/set —
+ * hence its snapshot record — can land after the shell has quiesced. We collect
+ * those write promises and await them here so the written value is pinned. Kept
+ * short: a pathological slow write must never stall the background capture; a key
+ * that does not settle in time is simply left unpinned (it drifts, the
+ * pre-snapshot behavior) rather than hanging. Reads that HIT are recorded
+ * synchronously during the render and do not depend on this.
+ */
+const SHELL_SNAPSHOT_WRITE_SETTLE_MS = 1000;
+
+/**
+ * Upper bound on the pre-render WRITE BARRIER: before the capture's match/render,
+ * settle the background tasks the FOREGROUND request already scheduled — its
+ * deferred ring-3 cacheRoute and ring-1 setItem writes all go through
+ * reqCtx.waitUntil, and every one of them is scheduled BEFORE scheduleShellCapture
+ * runs (the response, and its onResponse callbacks, are committed first). Draining
+ * them turns the capture's cache reads from a RACE into an ORDERING EDGE: the
+ * capture deterministically observes the foreground's cache generation, replays it
+ * (handler skipped, module-level side effects untouched), and records THAT
+ * generation into the snapshot — so prelude, snapshot, and ring-3 all agree on the
+ * foreground's generation and the capture can never clobber a foreground-produced
+ * entry with a re-render of its own. Scar tissue: without this, the capture's
+ * ring-3 lookup could land between the foreground write chain's serialization and
+ * its store.set, MISS, re-execute the route handler (bumping module-level
+ * counters), and — via the synthetic onResponse fire below — overwrite the
+ * foreground's entry (the mini shell-manifest regression). Bounded: a slow
+ * consumer waitUntil task must never stall the background capture; on timeout the
+ * capture proceeds with the pre-barrier (racy) behavior.
+ */
+const SHELL_CAPTURE_WRITE_BARRIER_MS = 1500;
+
+/**
+ * Settle the tracked background tasks on `reqCtx._pendingBackgroundTasks`,
+ * ITERATIVELY: a settled task can have scheduled a nested one (cache-store's
+ * cacheRoute outer task schedules the actual store.set in a second waitUntil), so
+ * each awaited batch may append more. Loop until no new tasks appear or the
+ * deadline passes. The capture's own task never enters the list
+ * (UNTRACKED_BACKGROUND_TASK), so the loop terminates.
+ */
+async function settleTrackedBackgroundTasks(
+  reqCtx: RequestContext<any>,
+  timeoutMs: number,
+): Promise<void> {
+  const tasks = reqCtx._pendingBackgroundTasks;
+  if (!tasks) return;
+  const deadline = Date.now() + timeoutMs;
+  let seen = 0;
+  while (tasks.length > seen) {
+    const remaining = deadline - Date.now();
+    if (remaining <= 0) return;
+    const batch = tasks.slice(seen);
+    seen = tasks.length;
+    let timer: ReturnType<typeof setTimeout> | undefined;
+    const guard = new Promise<void>((resolve) => {
+      timer = setTimeout(resolve, remaining);
+      (timer as { unref?: () => void }).unref?.();
+    });
+    await Promise.race([Promise.allSettled(batch).then(() => {}), guard]);
+    if (timer) clearTimeout(timer);
+  }
+}
+
+/**
+ * Delay before the in-place retry of a capture that produced no usable shell.
+ *
+ * The dominant reason a first capture comes back with a trivial prelude is a
+ * COLD render: in dev the module transform graph (route modules, the SSR/Flight
+ * transforms) is being built lazily and outlasts the task-quantized quiesce, so
+ * the shell has not finished rendering when we freeze it; on a cold worker the
+ * first invocation pays the same one-time cost. The first attempt WARMS that
+ * graph, so a second attempt a short beat later usually completes the shell in
+ * the SAME background task — no extra HTTP request needed. Short enough to feel
+ * instant, long enough for the module graph to settle. See
+ * docs/design/ppr-shell-resume.md ("Capture retry-in-place").
+ */
+const SHELL_CAPTURE_RETRY_DELAY_MS = 400;
+
+/** Sleep `ms`, unref'd so a Node dev process is never kept alive by the timer. */
+function delay(ms: number): Promise<void> {
+  return new Promise((resolve) => {
+    const t = setTimeout(resolve, ms);
+    (t as { unref?: () => void }).unref?.();
+  });
+}
+
+/**
+ * Module-level in-flight key set: the stampede guard for background captures, and
+ * its single owner. One capture runs per key per isolate; concurrent MISS/stale
+ * requests for the same key coalesce onto the first (the rest see the key present
+ * in scheduleShellCapture and skip). Added when a capture is scheduled and cleared
+ * in the task's finally once it settles, so a later request can recapture when TTL
+ * rolls. Living here (not split across the middleware) keeps the add/clear
+ * lifecycle in one layer.
+ */
+const inFlightCaptures = new Set<string>();
+
+/**
+ * Refused-capture backoff bounds. The window is EXPONENTIAL in the consecutive
+ * failure count: `min(BASE * 2^(failures-1), ceiling)` — 1s, 2s, 4s, … up to the
+ * mode's ceiling (60s in production, {@link REFUSED_CAPTURE_DEV_MAX_MS} in dev).
+ *
+ * Why exponential and not a flat 60s: a flat long window conflates two very
+ * different failures. A STRUCTURALLY ineligible route (no loading(), a cookie
+ * reader) fails forever and wants the long 60s cap. But a cold-but-ELIGIBLE route
+ * can also fail the in-place retry under a truly cold graph (dev module transform,
+ * or a cold worker under parallel load) — and it must recover FAST, on the next
+ * request or two, not be frozen for 60s (that would re-break the very cold-start DX
+ * the retry fixes; it bit the cloudflare dev e2e). Escalating from 1s means the
+ * eligible route re-probes almost immediately (warm now → HIT and clear), while the
+ * doomed route ramps to the ceiling within a handful of failures. Either way an
+ * app-wide mount never re-renders a doomed route on EVERY request.
+ */
+const REFUSED_CAPTURE_BASE_MS = 1_000;
+const REFUSED_CAPTURE_MAX_MS = 60_000;
+
+/**
+ * DEV-only backoff ceiling. In dev the 60s production cap is pure harm: the
+ * dominant no-shell cause is a COLD module graph (route modules, SSR/Flight
+ * transforms built lazily), and the very attempt that failed WARMS that graph, so
+ * the next attempt a beat later usually completes the shell. Capping the dev window
+ * low keeps a cold-but-eligible route re-probing every ~2s instead of freezing for
+ * up to 60s once the exponential climbs (1s→2s→4s→…→60s). A 60s freeze outlasts the
+ * e2e warm windows on cold CI runners: the capture races an unfinished shell,
+ * escalates the backoff past the poll window, and every subsequent request inside
+ * that window is skipped as backed-off — an eternal MISS for the test even though
+ * the modules are warm by then. Production keeps the full 60s cap: there the
+ * no-shell cause is far more likely to be a genuinely ineligible route (no
+ * loading()), which SHOULD be re-probed rarely. See #652 (item 3) and
+ * docs/design/ppr-shell-resume.md ("Refused-capture backoff").
+ */
+const REFUSED_CAPTURE_DEV_MAX_MS = 2_000;
+
+/**
+ * Dev signal, matching the rest of the RSC runtime (handler.ts, server-action.ts,
+ * progressive-enhancement.ts): treat anything but an explicit production build as
+ * dev. The build folds `process.env.NODE_ENV` to a literal, so this is a compile-
+ * time constant in the shipped worker — no runtime probe.
+ */
+function isDevMode(): boolean {
+  return process.env.NODE_ENV !== "production";
+}
+
+/** The active backoff ceiling for the current mode (dev capped low, prod at 60s). */
+function refusedCaptureCeilingMs(): number {
+  return isDevMode() ? REFUSED_CAPTURE_DEV_MAX_MS : REFUSED_CAPTURE_MAX_MS;
+}
+
+/**
+ * Refused-capture backoff: key -> { consecutive failure count, epoch ms until which
+ * the key is not re-probed }. A key enters backoff only after runShellCapture's
+ * in-place retry ALSO failed (or a genuine error). A successful capture clears the
+ * entry outright (failure count resets). Module-level (same lifetime as
+ * inFlightCaptures) so the whole lifecycle lives in one layer.
+ */
+const refusedCaptures = new Map<string, { failures: number; until: number }>();
+
+/** True iff `key` is still inside its (exponential) backoff window. */
+function isCaptureBackedOff(key: string): boolean {
+  const entry = refusedCaptures.get(key);
+  if (entry === undefined) return false;
+  // Window elapsed: allow a re-probe. Keep the entry (its failure count drives the
+  // NEXT window's escalation if the re-probe also fails); a success clears it.
+  return Date.now() < entry.until;
+}
+
+/**
+ * Record a refused/failed capture, escalating the backoff window exponentially up
+ * to the current mode's ceiling. The failure count keeps climbing across attempts
+ * (so a genuinely doomed route still ramps toward its cap), but the WINDOW is
+ * clamped: 60s in production, {@link REFUSED_CAPTURE_DEV_MAX_MS} in dev so a
+ * cold-but-eligible route re-probes fast instead of freezing out the e2e warm
+ * window on a cold CI runner (#652 item 3).
+ */
+function markCaptureBackoff(key: string): void {
+  const failures = (refusedCaptures.get(key)?.failures ?? 0) + 1;
+  const window = Math.min(
+    REFUSED_CAPTURE_BASE_MS * 2 ** (failures - 1),
+    refusedCaptureCeilingMs(),
+  );
+  refusedCaptures.set(key, { failures, until: Date.now() + window });
+}
+
+/** Clear any backoff for a key that just captured successfully. */
+function clearCaptureBackoff(key: string): void {
+  refusedCaptures.delete(key);
+}
+
+/**
+ * Keys already warned about a refused (null) capture, so the eternal-MISS shape
+ * logs once per key per isolate instead of on every request.
+ */
+const warnedNullCaptures = new Set<string>();
+
+/**
+ * Warn once per key that a capture produced no usable shell EVEN AFTER the
+ * in-place retry (runShellCapture attempt 2). Naming both causes with the
+ * distinguishing signal — does the route ever flip to HIT — is the whole point:
+ * the pre-retry version blamed "a loader route without loading()" unconditionally
+ * and misled users whose route DID have loading() and was merely cold. Because the
+ * retry already absorbs the cold-start case, by the time this fires cold-start has
+ * usually healed, so a firing warning leans toward the structural cause — but we
+ * still name both so a cold-start straggler is not misdiagnosed.
+ *
+ * The pointer is shipped-path-safe (a05c8251 convention): the /ppr skill ships in
+ * the npm tarball, but docs/design/ is repo-only, so link it by absolute GitHub URL
+ * rather than a relative path that dead-ends for consumers.
+ */
+function warnNullCaptureOnce(key: string): void {
+  if (warnedNullCaptures.has(key)) return;
+  warnedNullCaptures.add(key);
+  console.warn(
+    `[rango] Shell capture for "${key}" produced no usable shell after an in-place ` +
+      "retry; nothing was stored, so this request stays on MISS. Causes, told apart " +
+      "by whether the route ever flips to HIT:\n" +
+      "  1. Cold-start warmup (dev module transform, or a cold worker): the capture raced " +
+      "an unfinished shell render. This SELF-HEALS — the route flips to HIT once a later " +
+      "request warms the modules. Usually nothing to do.\n" +
+      "  2. Something suspends above <body> with no Suspense boundary and never settles " +
+      "within the capture window: a slower-than-the-capture-guard bake-lane loader " +
+      "(loaders on entries WITHOUT loading() execute at capture and their containers " +
+      "bake — the boundary-less await must settle for a shell to exist), or a pending " +
+      "promise consumed without a <Suspense> above it. The boundary belongs on the " +
+      "entry/component that OWNS the data: loading() on the entry that registers the " +
+      "loader (a child route's loading() does not unpin a parent layout's loaders), or " +
+      "a <Suspense> above the consuming component.\n" +
+      'See the /ppr skill (node_modules/@rangojs/router/skills/ppr/SKILL.md), "The hole ' +
+      'doctrine" and "The layout-with-loaders playbook", or the design docs: ' +
+      "https://github.com/ivogt/vite-rsc/blob/main/packages/rangojs-router/docs/design/ppr-shell-resume.md",
+  );
+}
+
+/** Keys already warned about a deterministic capture refusal (once per key). */
+const warnedRefusedCaptures = new Set<string>();
+
+/**
+ * Warn once per key that the capture was REFUSED for a deterministic reason
+ * (identity-guard trip or a rejected bake-lane loader). Distinct from
+ * warnNullCaptureOnce: these are not cold-start shapes, the retry is skipped,
+ * and the message carries the concrete cause instead of a differential.
+ */
+function warnCaptureRefusedOnce(key: string, reason: string): void {
+  if (warnedRefusedCaptures.has(key)) return;
+  warnedRefusedCaptures.add(key);
+  console.warn(
+    `[rango] Shell capture for "${key}" was refused: ${reason}\n` +
+      "The route stays on MISS (axis 1) — the page keeps working, only the shell " +
+      "cache is off. See the /ppr skill " +
+      "(node_modules/@rangojs/router/skills/ppr/SKILL.md) and " +
+      "docs/design/loader-container-bake.md.",
+  );
+}
+
+/** Keys already warned about untagged bake-lane data baked into the shell. */
+const warnedUntaggedShellBakes = new Set<string>();
+
+/**
+ * Warn once per shell key that a bake-lane loader baked material into the shell
+ * but the capture recorded ZERO tags (no render-collected _requestTags, no
+ * static ppr.tags). Such data is frozen in the shared shell until TTL and is
+ * un-evictable by tag: a server action refreshes the CLIENT only (rotates Rango
+ * state, busts the browser HTTP cache) and never touches the server shell store,
+ * and updateTag()/revalidateTag() cannot drop data that was baked WITHOUT a tag.
+ *
+ * Coarse per-shell-key signal, not per-loader attribution — the tag set is
+ * unioned globally at the write barrier, not tracked per loader, so we cannot
+ * name the offending loader without adding attribution plumbing (deliberately
+ * not done). Dev-only + once-per-key so it never spams production or fires on
+ * every capture.
+ */
+function warnUntaggedShellBakeOnce(key: string): void {
+  if (warnedUntaggedShellBakes.has(key)) return;
+  warnedUntaggedShellBakes.add(key);
+  console.warn(
+    `[rango] Shell capture for "${key}" baked bake-lane loader data into the ` +
+      "shell with NO cache tag. That data is frozen in the shared shell until " +
+      "TTL and cannot be tag-invalidated: a server action refresh touches the " +
+      "client only (not the server shell store), and updateTag() cannot evict " +
+      "data that was baked without a tag.\n" +
+      'Fix: tag the data (cacheTag() / "use cache" / cache({ tags })) so ' +
+      "updateTag() drops the shell, or move the volatile read under a loading() " +
+      "hole so it stays on the live lane and is never baked. See the /ppr skill " +
+      "(node_modules/@rangojs/router/skills/ppr/SKILL.md).",
+  );
+}
+
+/**
+ * Default cap (serialized UTF-8 bytes) on the capture data snapshot riding
+ * inside a shell entry, when the route's `ppr` option does not set
+ * `maxSnapshotBytes`. 8 MiB: the snapshot shares the stored envelope with the
+ * base64 prelude and the postponed blob, and the tightest store value limit is
+ * Cloudflare KV's 25 MiB — 8 MiB of snapshot leaves the envelope well under it
+ * while still fitting any sane pinned-ring payload. Applied ONLY in
+ * captureAndStoreShell (the single defaulting site — resolvePprConfig passes
+ * the option through undefaulted), so every producer and direct caller gets
+ * the same policy. Over the cap the snapshot is skipped (shell still stored;
+ * pinned reads drift — see PartialPrerenderProps.maxSnapshotBytes).
+ */
+export const DEFAULT_PPR_MAX_SNAPSHOT_BYTES: number = 8 * 1024 * 1024;
+
+/** Cached encoder for the snapshot byte measurement (one per module, not per capture). */
+const SNAPSHOT_BYTE_ENCODER = new TextEncoder();
+
+/** Keys already warned about an over-cap snapshot (once per key per isolate). */
+const warnedOverCapSnapshots = new Set<string>();
+
+/**
+ * Warn once per key that the capture data snapshot exceeded the route's
+ * `maxSnapshotBytes` cap and was skipped. The shell entry is still stored and
+ * served — only the pinned-read replay is lost, so shell-baked cached content
+ * can drift from the frozen prelude between capture and HIT and hydration
+ * repairs it client-side (the pre-snapshot behavior). Once per key: the same
+ * page recaptures on every TTL roll and would otherwise re-warn forever.
+ */
+function warnSnapshotOverCapOnce(
+  key: string,
+  snapshotBytes: number,
+  capBytes: number,
+): void {
+  if (warnedOverCapSnapshots.has(key)) return;
+  warnedOverCapSnapshots.add(key);
+  console.warn(
+    `[rango] Shell capture for "${key}" recorded a ${snapshotBytes}-byte data ` +
+      `snapshot, over the ${capBytes}-byte cap — the snapshot was skipped and ` +
+      "the shell was stored without it. The page keeps serving, but cached " +
+      "content baked into the shell is no longer pinned: if it drifts before " +
+      "the shell's TTL, hydration repairs the mismatch client-side. Raise the " +
+      "cap via the route's ppr option ({ maxSnapshotBytes }) if the entry " +
+      "still fits your store's value limit (Cloudflare KV: 25 MiB per value), " +
+      "or shrink the cache()'d data the shell bakes.",
+  );
+}
+
+/**
+ * One structured event from the background capture pipeline, mirroring the
+ * CFCacheReadDebugEvent pattern (cache/cf/cf-cache-types.ts): typed fields an
+ * operator can assert against, emitted per attempt and per skip, so the
+ * stored / no-shell / refused / backed-off lifecycle is observable outside
+ * dev console warnings. Configured via `createRouter({ debugShellCapture })`.
+ */
+export interface ShellCaptureDebugEvent {
+  /** Shell cache key the event is about. */
+  key: string;
+  /**
+   * What happened:
+   * - stored / redirect / no-shell / refused: one capture ATTEMPT's outcome
+   *   (see CaptureAttemptOutcome for the semantics of each)
+   * - error: the capture task failed with a genuine error (also routed through
+   *   reportCacheError; the key is backed off)
+   * - skip-in-flight: scheduleShellCapture found a capture already running for
+   *   the key (stampede guard) and scheduled nothing
+   * - skip-backoff: the key is inside its refused-capture backoff window and
+   *   the capture was not attempted
+   * - backoff: the key entered (or escalated) backoff after a terminal
+   *   no-shell — carries the new backoff state
+   */
+  outcome:
+    | "stored"
+    | "redirect"
+    | "no-shell"
+    | "refused"
+    | "error"
+    | "skip-in-flight"
+    | "skip-backoff"
+    | "backoff";
+  /** Attempt number (1 = first, 2 = in-place retry). Absent on skips. */
+  attempt?: number;
+  /** Wall-clock ms of the whole attempt (barrier + render + drain + put). */
+  attemptMs?: number;
+  /**
+   * Wall-clock ms the pre-render WRITE BARRIER waited on the foreground
+   * request's deferred cache writes (bounded by SHELL_CAPTURE_WRITE_BARRIER_MS).
+   */
+  barrierWaitMs?: number;
+  /**
+   * Wall-clock ms spent awaiting the capture's own deferred cache writes
+   * before the snapshot drain (bounded by SHELL_SNAPSHOT_WRITE_SETTLE_MS).
+   */
+  writeSettleMs?: number;
+  /** Stored prelude size in bytes (pre-base64). */
+  preludeBytes?: number;
+  /** Serialized snapshot size in UTF-8 bytes. Absent when nothing was recorded. */
+  snapshotBytes?: number;
+  /** True when the snapshot exceeded maxSnapshotBytes and was dropped. */
+  snapshotSkipped?: boolean;
+  /** Consecutive failure count in the key's backoff entry, when one exists. */
+  backoffFailures?: number;
+  /** Ms remaining in the key's backoff window, when one exists. */
+  backoffRemainingMs?: number;
+}
+
+/**
+ * Debug sink for the capture pipeline, mirroring {@link CFCacheDebug}: `true`
+ * logs each event to console (visible via `wrangler tail`), a function
+ * receives the events for programmatic capture. Off by default.
+ */
+export type ShellCaptureDebug =
+  | boolean
+  | ((event: ShellCaptureDebugEvent) => void);
+
+/**
+ * Compact single-line form of an event's fields, shared by the console sink
+ * and the dev Server-Timing mirror's `desc` (rsc-rendering). Plain
+ * alphanumerics/`=`/`-`/`()` only, so it needs no quoted-string escaping.
+ */
+export function describeShellCaptureEvent(
+  event: ShellCaptureDebugEvent,
+): string {
+  const parts: string[] = [event.outcome];
+  if (event.attempt !== undefined) parts.push(`attempt=${event.attempt}`);
+  if (event.attemptMs !== undefined) parts.push(`${event.attemptMs}ms`);
+  if (event.barrierWaitMs !== undefined) {
+    parts.push(`barrier=${event.barrierWaitMs}ms`);
+  }
+  if (event.writeSettleMs !== undefined) {
+    parts.push(`write-settle=${event.writeSettleMs}ms`);
+  }
+  if (event.preludeBytes !== undefined) {
+    parts.push(`prelude=${event.preludeBytes}b`);
+  }
+  if (event.snapshotBytes !== undefined) {
+    parts.push(
+      `snapshot=${event.snapshotBytes}b${event.snapshotSkipped ? " (over cap, skipped)" : ""}`,
+    );
+  }
+  if (event.backoffFailures !== undefined) {
+    parts.push(`backoff-failures=${event.backoffFailures}`);
+  }
+  if (event.backoffRemainingMs !== undefined) {
+    parts.push(`backoff-remaining=${event.backoffRemainingMs}ms`);
+  }
+  return parts.join(" ");
+}
+
+/** The `debugShellCapture: true` console sink: one compact line per event. */
+function consoleCaptureDebugSink(event: ShellCaptureDebugEvent): void {
+  console.log(
+    `[ShellCache][debug] ${event.key} ${describeShellCaptureEvent(event)}`,
+  );
+}
+
+/**
+ * Resolve the `debugShellCapture` router option to a callable sink, or
+ * undefined when off. The INTERNAL_RANGO_DEBUG env-flag fallback lives HERE
+ * (not at a call site) so every producer that resolves a sink inherits it;
+ * an explicit `false` wins over the env flag.
+ */
+export function resolveShellCaptureDebugSink(
+  option: ShellCaptureDebug | undefined,
+): ((event: ShellCaptureDebugEvent) => void) | undefined {
+  if (option === false) return undefined;
+  if (option === true) return consoleCaptureDebugSink;
+  if (typeof option === "function") return option;
+  return INTERNAL_RANGO_DEBUG ? consoleCaptureDebugSink : undefined;
+}
+
+/**
+ * Attempt-terminal outcomes recorded for the dev Server-Timing mirror. Skip
+ * events are excluded so a later request's skip cannot overwrite the
+ * interesting terminal event before a metrics-enabled request reads it.
+ */
+const TIMING_RECORDED_OUTCOMES = new Set<ShellCaptureDebugEvent["outcome"]>([
+  "stored",
+  "redirect",
+  "no-shell",
+  "refused",
+  "error",
+]);
+
+/**
+ * Dev-only last-terminal-event-per-key buffer backing the Server-Timing
+ * mirror: the capture runs AFTER its triggering response is committed, so its
+ * outcome can only ride a LATER response's header. rsc-rendering consumes this
+ * on the next ppr GET for the key when the metrics store is active
+ * (debugPerformance) and appends a `ppr:capture` Server-Timing entry. Dev-only
+ * (isDevMode) so production isolates never grow the map; FIFO-capped because
+ * with debugPerformance OFF nothing ever drains it, and a long dev session
+ * sweeping many URLs would otherwise accumulate one entry per shell key
+ * forever.
+ */
+const lastCaptureEventsForTiming = new Map<string, ShellCaptureDebugEvent>();
+const MAX_TIMING_EVENT_KEYS = 100;
+
+/**
+ * Consume (read-and-clear) the buffered terminal capture event for `key`, so
+ * one capture reports into exactly one later response's Server-Timing.
+ */
+export function takeCaptureDebugEventForTiming(
+  key: string,
+): ShellCaptureDebugEvent | undefined {
+  const event = lastCaptureEventsForTiming.get(key);
+  if (event) lastCaptureEventsForTiming.delete(key);
+  return event;
+}
+
+/**
+ * Publish one capture debug event: buffer terminal outcomes for the dev
+ * Server-Timing mirror, then hand the event to the configured sink. A
+ * throwing sink is swallowed — diagnostics must never fail a capture.
+ */
+function publishCaptureDebugEvent(
+  descriptor: Pick<ShellCaptureDescriptor, "debugSink">,
+  event: ShellCaptureDebugEvent,
+): void {
+  if (isDevMode() && TIMING_RECORDED_OUTCOMES.has(event.outcome)) {
+    // Refresh insertion order for the FIFO cap, then evict the oldest key.
+    lastCaptureEventsForTiming.delete(event.key);
+    if (lastCaptureEventsForTiming.size >= MAX_TIMING_EVENT_KEYS) {
+      const oldest = lastCaptureEventsForTiming.keys().next().value;
+      if (oldest !== undefined) lastCaptureEventsForTiming.delete(oldest);
+    }
+    lastCaptureEventsForTiming.set(event.key, event);
+  }
+  const sink = descriptor.debugSink;
+  if (!sink) return;
+  try {
+    sink(event);
+  } catch {
+    // Diagnostics only: a throwing consumer sink must never fail the capture.
+  }
+}
+
+/** Current backoff state fields for `key` (empty when no backoff entry). */
+function backoffFields(
+  key: string,
+): Pick<ShellCaptureDebugEvent, "backoffFailures" | "backoffRemainingMs"> {
+  const entry = refusedCaptures.get(key);
+  if (!entry) return {};
+  return {
+    backoffFailures: entry.failures,
+    backoffRemainingMs: Math.max(0, entry.until - Date.now()),
+  };
+}
+
+export interface FlightCaptureGate {
+  /** Identity passthrough of the source stream; feed this to captureShellHTML. */
+  stream: ReadableStream<Uint8Array>;
+  /**
+   * Resolves once the source has been byte-quiet for FLIGHT_QUIET_HOPS macrotask
+   * hops (or has closed — the DATA variant). At that instant the gate FREEZES:
+   * no further source byte reaches the fizz side, and the readable is left open
+   * (never closed / errored) so fizz postpones the still-pending references
+   * instead of seeing "Connection closed".
+   */
+  quiesce: Promise<void>;
+  /**
+   * Stop the internal macrotask-hop loop. captureShellHTML's maxWaitMs bounds the
+   * overall wait; dispose() is the clean shutdown for the pathological case where
+   * the source never goes byte-quiet (quiesce never fires), so the hop loop would
+   * otherwise keep rescheduling after captureShellHTML has already aborted and
+   * returned.
+   */
+  dispose(): void;
+}
+
+/**
+ * Wrap the capture Flight stream so the fizz shell prerender reads a stream that
+ * (a) forwards the shell rows unchanged, (b) resolves `quiesce` after the rows go
+ * byte-silent for FLIGHT_QUIET_HOPS macrotask hops, and (c) FREEZES at that
+ * instant — dropping any later byte without closing or erroring the readable, so
+ * the pending masked-loader references stay pending and fizz postpones them (the
+ * "unclosing stream" property, here for free because the masked rows never emit).
+ * Freezing also guarantees no post-quiesce byte — including an error row from any
+ * later abort/cancel of the underlying render — can corrupt the frozen prelude.
+ *
+ * Quiet is measured in TASKS, not wall-clock: after the first byte a macrotask
+ * hop loop compares a byte counter each turn and fires after K quiet turns. The
+ * hop timers are unref'd so they never keep a Node process alive, and the source
+ * closing (no holes) fires quiesce immediately for the DATA variant — the
+ * TransformStream then closes the readable, so fizz completes with postponed null.
+ *
+ * `holdUntil` keeps the gate from FREEZING before shell material with real latency
+ * has emitted. The hole doctrine bakes TOP-LEVEL pushed handle promises into the
+ * shell (resolvedHandleStream awaits them before the handles row emits), but a
+ * pushed promise that takes longer than the quiet window would otherwise be frozen
+ * out — the handles row would never reach fizz and the prelude would come back
+ * trivial. While `holdUntil` is pending, byte-quiet detection keeps running but the
+ * gate neither fires nor freezes; once it resolves, the quiet counter restarts so a
+ * burst of rows unblocked by it (the resolved handles row) is still captured. It
+ * never delays a HOLE from postponing: holes are pending promises that emit no
+ * bytes, so holding the gate open longer only ever admits shell rows. Bounded by
+ * captureShellHTML's maxWaitMs like every other quiesce input.
+ */
+export function gateFlightForCapture(
+  source: ReadableStream<Uint8Array>,
+  quietHops: number = FLIGHT_QUIET_HOPS,
+  holdUntil?: Promise<unknown>,
+): FlightCaptureGate {
+  let resolveQuiet!: () => void;
+  const quiesce = new Promise<void>((resolve) => {
+    resolveQuiet = resolve;
+  });
+
+  let bytesSeen = 0;
+  let armed = false;
+  let settled = false;
+  let disposed = false;
+  let frozen = false;
+  let held = holdUntil !== undefined;
+  let heldFirePending = false;
+
+  if (holdUntil !== undefined) {
+    const release = (): void => {
+      held = false;
+      if (heldFirePending && !settled && !disposed) {
+        // Quiet elapsed while held: restart the quiet count instead of firing
+        // immediately, so rows unblocked by the hold (the baked handles row)
+        // still flow before the freeze.
+        heldFirePending = false;
+        armed = false;
+        arm();
+      }
+    };
+    // Resolve OR reject releases the hold (a rejected handle value is dropped by
+    // resolveDeferredHandleValues; the capture must not hang on it).
+    holdUntil.then(release, release);
+  }
+
+  const fire = (): void => {
+    if (settled) return;
+    if (held) {
+      heldFirePending = true;
+      return;
+    }
+    settled = true;
+    frozen = true;
+    resolveQuiet();
+  };
+
+  const scheduleHop = (fn: () => void): void => {
+    const t = setTimeout(fn, 0);
+    // Never let the quiet-detection hop alone keep a Node process alive
+    // (no-op on workerd).
+    (t as { unref?: () => void }).unref?.();
+  };
+
+  // The hop loop starts only after the first byte, so it can never declare
+  // quiesce before fizz has begun pulling rows through the transform.
+  const arm = (): void => {
+    if (armed || settled || disposed) return;
+    armed = true;
+    let lastSeen = bytesSeen;
+    let quiet = 0;
+    const hop = (): void => {
+      if (settled || disposed) return;
+      if (bytesSeen === lastSeen) {
+        quiet += 1;
+        if (quiet >= quietHops) {
+          fire();
+          return;
+        }
+      } else {
+        lastSeen = bytesSeen;
+        quiet = 0;
+      }
+      scheduleHop(hop);
+    };
+    scheduleHop(hop);
+  };
+
+  const monitor = new TransformStream<Uint8Array, Uint8Array>({
+    transform(chunk, controller) {
+      // Post-quiesce: drop the byte. Do NOT enqueue and do NOT close/error — the
+      // frozen fizz input must stay a fixed byte set behind an open (unclosing)
+      // readable so still-pending references postpone.
+      if (frozen) return;
+      bytesSeen += chunk.length;
+      arm();
+      controller.enqueue(chunk);
+    },
+    flush() {
+      // Source closed with no freeze => DATA variant (no holes): quiet
+      // immediately. The TransformStream then closes the readable, so fizz
+      // completes and postponed comes back null.
+      fire();
+    },
+  });
+
+  return {
+    stream: source.pipeThrough(monitor),
+    quiesce,
+    dispose(): void {
+      disposed = true;
+    },
+  };
+}
+
+/**
+ * The background shell-capture descriptor: everything the capture task needs to
+ * store the shell. Built by the integrated PPR serve path (rsc-rendering.ts) from
+ * the route's `ppr` path option (`PartialPrerenderProps`) and the app-level cache
+ * store, and passed to scheduleShellCapture directly — it is NOT threaded through
+ * the request context. `tags` carries the route's OPERATIONAL `ppr.tags`; the
+ * capture UNIONS them with the shell's own auto-collected (non-loader) request
+ * tags from its derived render (the collected set stays authoritative). That
+ * union happens at the putShell WRITE BARRIER in captureAndStoreShell — after the
+ * capture quiesces — not at stream construction, so a tag recorded after an await
+ * in async shell content is still collected (issue #676). `store` is the same
+ * store the serve path resolved for its getShell read (requestCtx._cacheStore),
+ * so the capture writes where the serve reads.
+ */
+export interface ShellCaptureDescriptor {
+  key: string;
+  /**
+   * The RSC handler's build version (HandlerContext.version), stamped into the
+   * stored entry as ShellCacheEntry.buildVersion — the serve-side
+   * isValidShellHit gate compares it against the running build so a persistent
+   * store can never resume a stale build's postponed blob.
+   */
+  buildVersion: string;
+  ttl?: number;
+  swr?: number;
+  tags?: string[];
+  /**
+   * Per-route capture settle budget in ms (`ppr.captureTimeout`, resolved by
+   * resolvePprConfig). Feeds captureShellHTML's maxWaitMs — the ONE deadline
+   * bounding the whole capture, so it covers BOTH the fizz prerender AND the
+   * deferred-material settle window (the handlesBaked/loader-container
+   * holdUntil that keeps the gate from freezing while top-level pushes are
+   * pending). Undefined = SHELL_CAPTURE_MAX_WAIT_MS (15_000).
+   */
+  captureTimeout?: number;
+  store?: SegmentCacheStore<any>;
+  /** Gates the concise per-attempt capture breadcrumbs (INTERNAL_RANGO_DEBUG). */
+  debug?: boolean;
+  /**
+   * Cap (serialized UTF-8 bytes) on the entry's capture data snapshot; over it
+   * the snapshot is skipped and the shell stored without it (reported once per
+   * key). Absent = DEFAULT_PPR_MAX_SNAPSHOT_BYTES, applied in
+   * captureAndStoreShell — the single defaulting site.
+   */
+  maxSnapshotBytes?: number;
+  /**
+   * Structured capture-pipeline debug sink, resolved from
+   * `createRouter({ debugShellCapture })` (or INTERNAL_RANGO_DEBUG) via
+   * {@link resolveShellCaptureDebugSink}. Receives one
+   * {@link ShellCaptureDebugEvent} per attempt/skip.
+   */
+  debugSink?: (event: ShellCaptureDebugEvent) => void;
+}
+
+/**
+ * Schedule the background shell capture for a served document. Stampede-guarded:
+ * one capture per key per isolate. Runs via runBackground (waitUntil on workerd,
+ * fire-and-forget in Node dev), so the served response is never blocked on it. Any
+ * error is routed through reportCacheError — capture is best-effort; a failure just
+ * means the next request recaptures.
+ *
+ * Eligibility (nonce/allReady/partial/status/strategy) is decided by the caller
+ * (rsc-rendering.ts maybeScheduleShellCapture); this function only owns the
+ * stampede guard and the background dispatch.
+ */
+export function scheduleShellCapture(
+  ctx: HandlerContext<any>,
+  request: Request,
+  env: any,
+  url: URL,
+  reqCtx: RequestContext<any>,
+  ssrModule: SSRModule,
+  descriptor: ShellCaptureDescriptor,
+): void {
+  const key = descriptor.key;
+  if (inFlightCaptures.has(key)) {
+    publishCaptureDebugEvent(descriptor, { key, outcome: "skip-in-flight" });
+    return;
+  }
+  // Refused/failed within the window → skip the doomed re-render (one probe per
+  // key per window per isolate). Expired entries self-evict inside the check.
+  if (isCaptureBackedOff(key)) {
+    publishCaptureDebugEvent(descriptor, {
+      key,
+      outcome: "skip-backoff",
+      ...backoffFields(key),
+    });
+    return;
+  }
+  inFlightCaptures.add(key);
+  const captureTask = async () => {
+    try {
+      const outcome = await runShellCapture(
+        ctx,
+        request,
+        env,
+        url,
+        reqCtx,
+        ssrModule,
+        descriptor,
+      );
+      // Update the negative cache off the terminal outcome. A stored shell clears
+      // any prior backoff; a `no-shell` (after the in-place retry) backs the key
+      // off so the next requests don't re-probe it. A `redirect` has no shell but
+      // is not a doomed render — leave the backoff untouched.
+      if (outcome === "stored") clearCaptureBackoff(key);
+      else if (outcome === "no-shell") {
+        markCaptureBackoff(key);
+        publishCaptureDebugEvent(descriptor, {
+          key,
+          outcome: "backoff",
+          ...backoffFields(key),
+        });
+      }
+    } catch (error) {
+      // Detached background task — pass reqCtx so onError still fires when the ALS
+      // context is gone. A genuine failure recurs, so back it off too (re-probe
+      // once per window, not every request) and report it once.
+      markCaptureBackoff(key);
+      publishCaptureDebugEvent(descriptor, {
+        key,
+        outcome: "error",
+        ...backoffFields(key),
+      });
+      reportCacheError(error, "cache-write", "[ShellCache] capture", reqCtx);
+    } finally {
+      inFlightCaptures.delete(key);
+    }
+  };
+  // Serialize capture EXECUTION per isolate (capture-queue.ts): concurrent
+  // captures starve each other's task-quantized quiet windows — one grinding
+  // capture makes the sibling freeze a trivial prelude and store nothing
+  // (rotating eternal-MISS victims on GH runners). The stampede guard above
+  // stays per-key (dedupe while queued); the queue is cross-key.
+  const serializedTask = () => enqueueSerializedCapture(captureTask);
+  // The capture's own task must NOT enter reqCtx._pendingBackgroundTasks: the
+  // capture drains that list before rendering (the write-barrier ordering edge),
+  // and awaiting its own still-running promise would burn the whole barrier
+  // deadline on every capture.
+  (serializedTask as { [UNTRACKED_BACKGROUND_TASK]?: boolean })[
+    UNTRACKED_BACKGROUND_TASK
+  ] = true;
+  runBackground(reqCtx, serializedTask);
+}
+
+/**
+ * The outcome of one capture attempt.
+ * - `stored`: a usable shell was captured (and a putShell was attempted; a store
+ *   I/O failure is reported separately and does NOT make the attempt retryable —
+ *   the capture itself worked).
+ * - `redirect`: the matched route redirects, so there is no shell to capture.
+ * - `no-shell`: the prelude came back trivial (no <body>) OR captureShellHTML
+ *   rejected with our own abort. This is the only RETRYABLE outcome.
+ */
+type CaptureAttemptOutcome = "stored" | "redirect" | "no-shell" | "refused";
+
+/**
+ * Per-attempt observability fields, filled along the capture path (barrier in
+ * attemptCapture, the rest in captureAndStoreShell) and folded into the
+ * attempt's {@link ShellCaptureDebugEvent} by runShellCapture. A plain mutable
+ * bag, not a return value: captureAndStoreShell's outcome type stays a string
+ * union its existing callers (producer B, tests) consume unchanged.
+ */
+interface CaptureAttemptStats {
+  barrierWaitMs?: number;
+  writeSettleMs?: number;
+  preludeBytes?: number;
+  snapshotBytes?: number;
+  snapshotSkipped?: boolean;
+}
+
+/**
+ * Run the shell capture with a single in-place retry, then store the result.
+ *
+ * Each attempt re-derives EVERYTHING (fresh context, fresh router.match, fresh
+ * Flight render) via {@link attemptCapture} — a capture consumes its handle store,
+ * its request-tag set, and its one-shot Flight stream, so none of them are
+ * reusable across attempts. A first attempt that comes back `no-shell` is almost
+ * always a cold render (dev module transform / cold worker) that had not finished
+ * when we froze the shell; the attempt itself warmed the module graph, so a second
+ * attempt a short beat later usually completes the shell in the SAME background
+ * task. That kills the old multi-request warmup where the caller had to re-issue
+ * several HTTP requests before a capture stuck. We retry ONLY on `no-shell` (and a
+ * defensively-caught abort); a genuine render error is NOT retried — it propagates
+ * to scheduleShellCapture's reportCacheError. See docs/design/ppr-shell-resume.md.
+ *
+ * `retryDelayMs` is a parameter (defaulting to the module const) so unit tests can
+ * drive the retry without a real 400ms wall-clock wait.
+ */
+async function runShellCapture(
+  ctx: HandlerContext<any>,
+  request: Request,
+  env: any,
+  url: URL,
+  reqCtx: RequestContext<any>,
+  ssrModule: SSRModule,
+  descriptor: ShellCaptureDescriptor,
+  retryDelayMs: number = SHELL_CAPTURE_RETRY_DELAY_MS,
+): Promise<CaptureAttemptOutcome> {
+  const log = descriptor.debug
+    ? (message: string) => console.log(message)
+    : () => {};
+
+  // One attempt + its structured debug event: the stats object rides through
+  // attemptCapture/captureAndStoreShell collecting the observability fields
+  // (barrier wait, write-settle wait, prelude/snapshot bytes), and the event
+  // folds them with the outcome. A genuine render error skips the attempt
+  // event — scheduleShellCapture's catch publishes the terminal `error` event.
+  const timedAttempt = async (
+    attempt: number,
+  ): Promise<CaptureAttemptOutcome> => {
+    const stats: CaptureAttemptStats = {};
+    const start = performance.now();
+    const outcome = await attemptCapture(
+      ctx,
+      request,
+      env,
+      url,
+      reqCtx,
+      ssrModule,
+      descriptor,
+      stats,
+    );
+    publishCaptureDebugEvent(descriptor, {
+      key: descriptor.key,
+      outcome,
+      attempt,
+      attemptMs: Math.round(performance.now() - start),
+      ...stats,
+    });
+    return outcome;
+  };
+
+  const first = await timedAttempt(1);
+  // "refused" is deterministic (identity guard / rejected bake-lane loader —
+  // its own warning already fired): no retry, and the caller backs the key off
+  // exactly like a structural no-shell.
+  if (first === "refused") return "no-shell";
+  // "stored" (success) or "redirect" (no shell exists): nothing to retry.
+  if (first !== "no-shell") return first;
+
+  // Attempt 1 produced no usable shell. Retry ONCE in place — the first attempt
+  // warmed the dev transform graph / cold worker, so attempt 2 typically completes
+  // the shell without another HTTP request. The concise line is gated on the
+  // middleware's debug flag (threaded via the descriptor) so it replaces the old
+  // full DOMException dump with one readable breadcrumb.
+  log(
+    `[ShellCache] capture attempt 1/2 for ${descriptor.key} aborted before shell completed (cold modules?) — retrying`,
+  );
+  await delay(retryDelayMs);
+  const second = await timedAttempt(2);
+  if (second === "refused") return "no-shell";
+  if (second !== "no-shell") return second;
+
+  // Both attempts came back with no usable shell. Cold-start would have healed by
+  // now, so the eternal-MISS structural shape (a loader route without loading()) is
+  // the likely cause — warn once per key. Ordering matters: because the retry
+  // absorbs cold-start, cold-start routes almost never reach this warning. The
+  // caller (scheduleShellCapture) reads this `no-shell` return to back the key off.
+  log(
+    `[ShellCache] capture attempt 2/2 for ${descriptor.key} aborted — giving up until next request`,
+  );
+  warnNullCaptureOnce(descriptor.key);
+  return "no-shell";
+}
+
+/** Fold the capture's handle-liveness record into the entry flag (true | undefined). */
+function handlerLayerIsLive(
+  liveness: RequestContext["_shellCaptureHandleLiveness"],
+): true | undefined {
+  if (!liveness) return undefined;
+  return liveness.holes ||
+    liveness.pendingPushes > 0 ||
+    liveness.handlerInvokedLoader
+    ? true
+    : undefined;
+}
+
+/**
+ * One capture attempt in a DERIVED request context.
+ *
+ * The derived context is `Object.create(reqCtx)` so it inherits the foreground's
+ * post-middleware state (variables, cache store, env/request/url, waitUntil) while
+ * overriding the render-scoped accumulators as own properties:
+ *   - _handleStore: a fresh store. The foreground store is already drained to
+ *     completion (its stream() flipped `completed` on settle) and would throw
+ *     LateHandlePushError on any re-push. Every downstream reader resolves the
+ *     store off the ambient context (setupLoaderAccess captures
+ *     _getRequestContext()._handleStore; trackHandler reads it), so the fresh
+ *     store on the derived context is what the capture match() writes handles to.
+ *   - _requestTags: a fresh Set. The capture collects its OWN shell tags here —
+ *     non-loader tags only, since loaders are masked — which is exactly the tag
+ *     set a shell entry should be invalidatable by (loader tags belong to holes).
+ *   - _transitionWhen: a fresh [] so the capture's transition gating is its own.
+ *   - _shellCaptureRun: true — the switch loaders/cookies/headers guards read.
+ *   - _metricsStore: undefined so the capture never appends to the foreground's
+ *     (already-finalized) metrics.
+ *   - _renderBarrier family: an own barrier wired to the fresh handle store
+ *     (wireRenderBarrier), plus _treeHasStreaming/deadlock-guard resets — the
+ *     capture's rendered() lifecycle is its own, not the foreground's.
+ *
+ * The capture is MIXED-CHAIN: its match() behaves like a normal render with
+ * respect to the segment cache — cache()'d segments replay from ring 3, UNCACHED
+ * segments execute their handlers fresh (which is why the cookies()/headers()
+ * capture guard is load-bearing). Middleware is NOT re-run: it already ran for the
+ * triggering request, and the derived context inherits its post-middleware state
+ * (guarding is serve-time; the shell is never served without the full chain).
+ *
+ * A FRESH context (and match/render) per attempt is what makes the retry sound:
+ * the second attempt is a clean capture, not a resumption of the first.
+ */
+async function attemptCapture(
+  ctx: HandlerContext<any>,
+  request: Request,
+  env: any,
+  url: URL,
+  reqCtx: RequestContext<any>,
+  ssrModule: SSRModule,
+  descriptor: ShellCaptureDescriptor,
+  stats: CaptureAttemptStats,
+): Promise<CaptureAttemptOutcome> {
+  // WRITE BARRIER (ordering edge, not a narrower race): settle the foreground's
+  // already-scheduled background tasks — its deferred ring-3/ring-1 cache writes —
+  // BEFORE this attempt's match/render, so the capture's cache reads observe the
+  // foreground's generation deterministically. Contract: a capture must never
+  // clobber a ring-3 entry the foreground produced; with the barrier, the
+  // capture's ring-3 lookup HITs the foreground's entry and REPLAYS it (handler
+  // skipped, cache-store middleware's write path gated off by state.cacheHit), so
+  // prelude, snapshot, and ring-3 agree on the foreground's generation. Runs per
+  // attempt (the retry re-checks; already-settled promises are free).
+  const barrierStart = performance.now();
+  await settleTrackedBackgroundTasks(reqCtx, SHELL_CAPTURE_WRITE_BARRIER_MS);
+  stats.barrierWaitMs = Math.round(performance.now() - barrierStart);
+
+  const { derivedCtx, freshHandleStore } = deriveShellCaptureContext(
+    reqCtx,
+    descriptor,
+  );
+
+  return runWithRequestContext(derivedCtx, async () => {
+    const match = await ctx.router.match(request, { env });
+    // A route that redirects has no shell to capture — bail (no store write, no
+    // retry: a redirect is deterministic).
+    if (match.redirect) return "redirect";
+
+    setRequestContextParams(match.params, match.routeName);
+
+    const payload = buildFullPayload(
+      match,
+      ctx,
+      url,
+      derivedCtx,
+      freshHandleStore,
+    );
+    const flightStage = renderRscFlightStage(
+      {
+        ctx,
+        request,
+        env,
+        url,
+        payload,
+        tracking: {
+          mode: "full",
+          routeKey: derivedCtx._routeName,
+          phases: RSC_FLIGHT_ONLY_PHASES,
+        },
+      },
+      performance.now(),
+    );
+
+    // Pass the descriptor with its STATIC ppr.tags unchanged. The shell's own
+    // render-recorded tags are snapshotted at the putShell WRITE BARRIER inside
+    // captureAndStoreShell, not here: a tag recorded AFTER an await in async shell
+    // content (and tags propagated by async cache()/"use cache" reads) lands after
+    // this synchronous construction point, so snapshotting here dropped it — the
+    // shell-tag snapshot must sit behind the quiesce gate (issue #676).
+    return captureAndStoreShell(
+      ssrModule,
+      flightStage.stream,
+      freshHandleStore,
+      derivedCtx,
+      descriptor,
+      stats,
+    );
+  });
+}
+
+/**
+ * The derived capture context and its fresh (mask-funneled) handle store,
+ * shared by BOTH shell producers: the runtime background capture
+ * (attemptCapture, producer A) and the build-time prerender shell capture
+ * (prerender/build-shell-capture.ts, producer B — issue #699). One
+ * implementation so the capture semantics — the nested-thenable mask funnel,
+ * handler-liveness bookkeeping, snapshot recording, the implicit doc-cache
+ * scope — cannot drift between producers.
+ */
+export interface CaptureContextDerivation {
+  derivedCtx: RequestContext;
+  freshHandleStore: HandleStore;
+}
+
+/**
+ * Derive the capture request context from a base context. Producer A passes
+ * the foreground request's post-middleware context (the derived context
+ * inherits its variables/env/cookie machinery through the prototype);
+ * producer B passes a synthetic build-request context created by
+ * createRequestContext over the build env, with a fresh MemorySegmentCacheStore
+ * as `_cacheStore` so the recording/snapshot machinery arms identically.
+ */
+export function deriveShellCaptureContext(
+  reqCtx: RequestContext<any>,
+  descriptor: Pick<ShellCaptureDescriptor, "ttl" | "swr">,
+): CaptureContextDerivation {
+  const freshHandleStore = createHandleStore();
+  freshHandleStore.onError = reqCtx._handleStore.onError;
+  // Shape = liveness for handles, exactly as for bake-lane loader containers
+  // (mask-nested.ts): nested thenables in a pushed handle container are
+  // per-request by declaration, so the CAPTURE's copy masks them — the
+  // consuming boundary postpones as a hole regardless of settle timing,
+  // instead of a fast-settling nested value baking into the shared shell. A
+  // TOP-LEVEL promise push keeps its documented bake contract (awaited
+  // pre-SSR, gate held open for it), but the container it RESOLVES to gets
+  // the same nested masking. Wrapping THIS store's push is the single funnel:
+  // the store exists only for this capture attempt, so every push wrapper
+  // (setupLoaderAccess, createUseFunction, prerender) inherits the policy and
+  // the foreground store is untouched.
+  // Shell fast path bookkeeping on the same funnel:
+  //  - handleLiveness: a nested thenable in a push made OUTSIDE a DSL loader
+  //    scope (attribution read synchronously at push time — handler bodies,
+  //    handler-invoked ctx.use(loader) callbacks, defers) declares
+  //    handler-layer per-request data. Its mask is a hole only a handler
+  //    re-run can fill, so the entry must not serve handler-free
+  //    (ShellCacheEntry.handlerLiveHoles). Still-pending top-level handler
+  //    pushes at the putShell barrier count too — their liveness is unknowable.
+  //  - loaderScopedPushValues: DSL-loader pushes re-run fresh on every HIT, so
+  //    their captured values must NOT enter a segment record's handle snapshot
+  //    (replay would duplicate the fresh push, and their masked nested
+  //    promises would stall the Flight handle encode to its timeout). The set
+  //    rides the derived context (_shellCaptureLoaderHandleValues) and is
+  //    applied ONLY at the captureHandles cache-write call site — every other
+  //    getDataForSegment consumer (the render-barrier snapshot, prerender)
+  //    sees every push.
+  const handleLiveness = {
+    holes: false,
+    pendingPushes: 0,
+    handlerInvokedLoader: false,
+  };
+  const loaderScopedPushValues = new WeakSet<object>();
+  const rawCapturePush = freshHandleStore.push.bind(freshHandleStore);
+  freshHandleStore.push = (
+    handleName: string,
+    segmentId: string,
+    value: unknown,
+  ) => {
+    const pushedInLoaderScope = isInsideLoaderScope();
+    // Single walk: the mask reports whether it masked any nested thenable
+    // (the liveness declaration) while building the capture copy.
+    const maskWithLiveness = (v: unknown): unknown => {
+      const report: MaskReport = { thenable: false };
+      const masked = maskNestedContainerThenables(v, undefined, report);
+      if (!pushedInLoaderScope && report.thenable) {
+        handleLiveness.holes = true;
+      }
+      return masked;
+    };
+    let masked: unknown;
+    if (isThenable(value)) {
+      if (!pushedInLoaderScope) {
+        handleLiveness.pendingPushes++;
+        const settle = () => handleLiveness.pendingPushes--;
+        value.then(settle, settle);
+      }
+      masked = value.then(maskWithLiveness);
+    } else {
+      masked = maskWithLiveness(value);
+    }
+    if (pushedInLoaderScope && typeof masked === "object" && masked !== null) {
+      loaderScopedPushValues.add(masked);
+    }
+    rawCapturePush(handleName, segmentId, masked);
+  };
+
+  const derivedCtx: RequestContext = Object.create(reqCtx);
+  derivedCtx._handleStore = freshHandleStore;
+  // Own render barrier, closure-bound to the derived ctx and the fresh store
+  // (issue #684, plan 009). Without this every _renderBarrier* read fell
+  // through the prototype to the foreground's ALREADY-RESOLVED barrier: a
+  // bake-lane loader's `await ctx.rendered()` resolved instantly and
+  // ctx.use(handle) read the FOREGROUND handle snapshot — foreground
+  // per-request handle data could bake into the shared shell. wireRenderBarrier
+  // also resets _treeHasStreaming (recomputed for the capture's tree) and the
+  // deadlock-guard fields as own properties.
+  wireRenderBarrier(derivedCtx, freshHandleStore);
+  derivedCtx._shellCaptureLoaderHandleValues = loaderScopedPushValues;
+  derivedCtx._requestTags = new Set<string>();
+  // Own explicit-store registry: cache-store resolutions during the capture
+  // (the implicit scope's SnapshotOnlySegmentStore, any per-capture explicit
+  // store instance) must NOT register into the handler-lifetime
+  // _explicitTaggedStores set — a capture-ephemeral store pinned there would
+  // trip the partial-tag-store warning on every later updateTag() and retain
+  // the whole capture snapshot in memory. Capture registrations die with this
+  // context; module-singleton stores stay registered by normal renders.
+  derivedCtx._explicitTaggedStores = new Set();
+  derivedCtx._transitionWhen = [];
+  derivedCtx._shellCaptureRun = true;
+  derivedCtx._metricsStore = undefined;
+  // Spans, like perf metrics above, are a FOREGROUND surface: the capture
+  // re-render must not emit a second rango.render/loader/ssr set after the
+  // foreground rango.request span ended (orphan spans in the trace).
+  // _tracing is otherwise inherited through Object.create(reqCtx).
+  derivedCtx._tracing = undefined;
+  // Bake-lane loader containers (loaders on entries with no renderable
+  // loading() execute during capture — docs/design/loader-container-bake.md).
+  // resolveLoaderData registers each container promise here; the drain in
+  // captureAndStoreShell elides + pins them into the snapshot's loader family.
+  derivedCtx._shellCaptureLoaderRecords = new Map();
+  // Own onResponse list so the capture's match-middleware callbacks (the ring-3
+  // segment cache write registers here) are ISOLATED from the foreground's shared
+  // array AND can be fired by captureAndStoreShell. The segment write is gated
+  // behind onResponse, which the capture never triggers (it builds no Response) —
+  // without firing it, a ring-3 cache() MISS at capture renders fresh into the
+  // prelude but is never written, so it is never recorded and drifts on a HIT.
+  derivedCtx._onResponseCallbacks = [];
+
+  // Capture data snapshot: read every cache-store hit/write through a recording
+  // wrapper on the DERIVED context's store (own property, so the shared
+  // reqCtx._cacheStore is untouched — the snapshot is per-capture). Its records
+  // ride inside the ShellCacheEntry so a HIT can reproduce the shell's cached
+  // content byte-identically. See cache/shell-snapshot.ts and the design doc.
+  //
+  // Cache writes are deferred (waitUntil): a MISS-at-capture value's setItem/set
+  // — hence its record — would otherwise land after the shell quiesces. Override
+  // the derived context's waitUntil to COLLECT those write promises (still
+  // forwarding to the parent so the write persists and the worker stays alive),
+  // then captureAndStoreShell awaits them before draining. Reads that HIT are
+  // recorded synchronously during the render and need none of this.
+  derivedCtx._shellCaptureHandleLiveness = handleLiveness;
+  if (reqCtx._cacheStore) {
+    const recordingStore = new RecordingShellStore(reqCtx._cacheStore);
+    derivedCtx._cacheStore = recordingStore;
+    derivedCtx.waitUntil = (fn: () => Promise<void>): void => {
+      const p = Promise.resolve().then(fn);
+      recordingStore.trackWrite(p);
+      reqCtx.waitUntil(() => p);
+    };
+    // Shell fast path (capture side): the implicit doc-cache scope makes the
+    // capture's match write ALL matched non-loader segments as one doc-keyed
+    // segment record — into the snapshot only (SnapshotOnlySegmentStore), so
+    // the record dies with the shell entry and the next capture's lookup
+    // still misses (handlers re-run on recapture). Routes deriving their own
+    // cache scope are untouched (resolveShellImplicitCacheScope).
+    derivedCtx._shellImplicitCache = {
+      ttl: descriptor.ttl,
+      swr: descriptor.swr,
+      store: new SnapshotOnlySegmentStore(recordingStore),
+    };
+  }
+
+  return { derivedCtx, freshHandleStore };
+}
+
+/**
+ * Seal handles, derive the quiesce signal, prerender + abort via the SSR module's
+ * captureShellHTML, and store the result. Returns the attempt outcome (the caller
+ * owns retry/warn decisions — this function no longer warns). Never throws out of
+ * the store write: a failed putShell is routed through reportCacheError so the
+ * background task stays best-effort, and the attempt still counts as `stored` (the
+ * capture worked; only the store I/O failed). `ssrModule.captureShellHTML` MUST be
+ * present (eligibility is checked before scheduling).
+ *
+ * A `no-shell` result (trivial prelude, or a defensively-caught abort) is the only
+ * retryable outcome; a genuine (non-abort) captureShellHTML error propagates so it
+ * reaches reportCacheError and is NOT retried.
+ */
+async function captureAndStoreShell(
+  ssrModule: SSRModule,
+  rscStream: ReadableStream<Uint8Array>,
+  handleStore: HandleStore,
+  reqCtx: RequestContext<any>,
+  capture: ShellCaptureDescriptor,
+  stats?: CaptureAttemptStats,
+): Promise<Exclude<CaptureAttemptOutcome, "redirect">> {
+  const captureShellHTML = ssrModule.captureShellHTML!;
+
+  // Seal the handle store so the payload's handles generator (resolvedHandleStream
+  // -> handleStore.stream()) converges and completes even though masked loaders
+  // never resolve. handleStore.settled gates ONLY on tracked HANDLER promises
+  // (handleStore.track, via trackHandler) — NOT on deferred handle VALUES pushed
+  // through ctx.use(Handle).defer(), which are plain pushed promises. So seal()
+  // does not reject or hang on outstanding defers: settled resolves once the
+  // handlers settle, and each deferred slot resolves on its own createDeferred
+  // timeout (defer.ts, default 10s) or when its resolver fires. A defer whose
+  // resolver depends on a masked loader can never fire, so it stays pending until
+  // that 10s timeout — longer than maxWaitMs (5s). At the abort the handles
+  // generator has not yielded, SsrRoot suspends at the root (consumeAsyncGenerator
+  // sits above every boundary), the prelude comes back trivial, and
+  // captureShellHTML's sanity gate returns null: the designed fail-safe no-op, not
+  // an error. This mirrors the __prerender_collect seal+settled regime, which also
+  // excludes loaders. See docs/design/ppr-shell-resume.md ("Loaders and handles").
+  handleStore.seal();
+
+  // Handles contract, shell half ("nesting = liveness"): TOP-LEVEL pushed handle
+  // promises are BAKED into the shell — resolvedHandleStream awaits them before
+  // the payload's handles row emits. A pushed promise with real latency would lose
+  // the byte-quiet race (the pending handles row emits no bytes, the gate freezes,
+  // the row is dropped, SsrRoot suspends at the root), so the gate is HELD open
+  // until the same await completes: handlesBaked mirrors resolvedHandleStream's
+  // resolution (getData waits the tracked-handler barrier; resolveDeferredHandleValues
+  // awaits the top-level thenables). NESTED promises inside pushed containers are
+  // shallow-skipped by isThenable and never hold the gate — they stay holes.
+  // Bounded by maxWaitMs like every quiesce input (a defer hanging on a masked
+  // loader still ends in the sanity-gate refusal).
+  const handlesBaked = handleStore.getData().then(resolveDeferredHandleValues);
+  // Bake-lane loader containers hold the gate the same way (loader-container-
+  // bake): a boundary-less container with real latency (a 100ms layout loader)
+  // would otherwise lose the 2-hop byte-quiet race — the pending loaderData row
+  // emits no bytes, the gate freezes, and the awaiting tree pins above <body>.
+  // The records map is fully populated before this point (loader promises are
+  // created during the capture's match()), so the hold covers every bake-lane
+  // container. allSettled: a REJECTED container releases the hold (the drain
+  // below refuses the capture); nested promises INSIDE a container never hold
+  // the gate — they stay holes. Bounded by maxWaitMs like every quiesce input.
+  const loaderRecordsForHold = reqCtx._shellCaptureLoaderRecords;
+  const holdUntil =
+    loaderRecordsForHold && loaderRecordsForHold.size > 0
+      ? Promise.allSettled([handlesBaked, ...loaderRecordsForHold.values()])
+      : handlesBaked;
+  const gate = gateFlightForCapture(rscStream, undefined, holdUntil);
+  // Quiesce = handles settled AND the Flight shell rows went task-quiet. Either
+  // half stalling is bounded by captureShellHTML's maxWaitMs.
+  const quiesce = Promise.all([handleStore.settled, gate.quiesce]).then(
+    () => {},
+  );
+
+  // Deterministic identity-guard refusal, checked at BOTH exits below: the
+  // guard error either rejects the prerender itself (boundary-less segment —
+  // lands in the catch) or is swallowed into per-loader error UI (the render
+  // completes — caught after the try). One helper so the message and the
+  // "refused" mapping cannot drift between the two sites.
+  const refuseOnGuardTrip = (): "refused" | undefined => {
+    const fnName = reqCtx._shellCaptureGuardTripped;
+    if (!fnName) return undefined;
+    // Name the recorded source instead of hardcoding a lane. Under the
+    // consumption-lane rule, handler-INVOKED loader bodies are exempt from
+    // the guard (their value is a baked shared copy, mirroring cache()), so a
+    // trip can only come from a bake-lane SEGMENT loader or from non-loader
+    // handler/render code — the old blanket "bake-lane loader" attribution
+    // sent a live-lane debugging session down the wrong lane (issue #672).
+    const loaderId = reqCtx._shellCaptureGuardTrippedLoaderId;
+    const origin = loaderId
+      ? `segment loader "${loaderId}"`
+      : "handler/render code (no loader body was executing)";
+    warnCaptureRefusedOnce(
+      capture.key,
+      `${origin} called ${fnName}() during capture. Identity must not bake into a shared shell. ` +
+        "For a segment loader (bake lane, no loading()): give its entry a loading() boundary " +
+        "(the live lane, masked at capture) or move the identity-dependent part into a nested " +
+        "promise. For handler/render code: keep the value live by consuming a loader " +
+        'client-side (useLoader in a "use client" component). Note: `await ctx.use(loader)` ' +
+        "inside a HANDLER is exempt from this guard — its value bakes into the shared shell " +
+        "as a capture-time copy, mirroring cache() semantics (the consumption-lane rule).",
+    );
+    return "refused";
+  };
+
+  try {
+    // captureShellHTML CONSUMES the (gated) stream — it is not also SSR'd.
+    let result: Awaited<ReturnType<typeof captureShellHTML>>;
+    try {
+      // One deadline for the whole capture — semantics spec'd on the option
+      // (PartialPrerenderProps.captureTimeout, urls/pattern-types.ts).
+      result = await observePhase(PHASES.ssr, () =>
+        captureShellHTML(gate.stream, {
+          quiesce,
+          maxWaitMs: capture.captureTimeout ?? SHELL_CAPTURE_MAX_WAIT_MS,
+        }),
+      );
+    } catch (error) {
+      // Guard-tripped rejection arrives here (not at the drain) — refuse
+      // BEFORE the AbortError-vs-rethrow decision below.
+      const refused = refuseOnGuardTrip();
+      if (refused) return refused;
+      // captureShellHTML normally converts its OWN deliberate abort to a null
+      // return (index.tsx). This catch is defensive: if an AbortError still escapes
+      // (a runtime where the abort surfaces as a stream rejection outside its
+      // guard), treat it as the same retryable "no usable shell" degradation rather
+      // than a failure — do NOT report it as an error. A genuine (non-abort) render
+      // error is a real failure: rethrow so it reaches reportCacheError (no retry).
+      if ((error as { name?: string } | null)?.name === "AbortError") {
+        return "no-shell";
+      }
+      throw error;
+    }
+
+    // null = sanity gate refused (trivial/empty prelude, no <body>). Store nothing
+    // and report `no-shell` so the caller (runShellCapture) can retry once and, if
+    // that also fails, warn once per key. On a cold render this is the shell not
+    // yet finished; on a loader route WITHOUT a route-level loading() boundary it is
+    // the structural eternal-MISS shape (the masked loader pins the tree above
+    // <body> at tree-build). The caller's warning names both.
+    // Guard check first — BEFORE the trivial-prelude retry path. A guard trip
+    // is deterministic (retrying re-trips it), and when the tripping loader's
+    // error UI still completed a shell, storing it would bake the failure into
+    // a shared page.
+    const refused = refuseOnGuardTrip();
+    if (refused) return refused;
+
+    if (result === null) {
+      return "no-shell";
+    }
+    if (stats) stats.preludeBytes = result.prelude.length;
+
+    // Store per the flag's key/ttl/swr/tags, into the flag's store: the middleware
+    // threads the SAME store it resolved for its getShell read (options.store ??
+    // _cacheStore), so a store-attached middleware writes captures where it reads
+    // them. The _cacheStore fallback covers a flag armed without a store (tests).
+    // reactVersion is read from the same React.version import the middleware
+    // validates reads against, so capture and serve always agree.
+    // Fire the capture's isolated onResponse callbacks with a synthetic 200 so
+    // the ring-3 segment cache write (cacheScope.cacheRoute, registered via
+    // onResponse by the cache-store match-middleware and gated on a 200) runs
+    // DURING capture, routed through the recording store. The foreground path
+    // never fires for the capture — it builds no Response — so without this a
+    // cache() SEGMENT that MISSED at capture would be rendered fresh into the
+    // prelude yet never written, hence never recorded, and would drift on a HIT
+    // (an item-family "use cache" write already runs inline during the render, so
+    // it needs none of this; only segment writes are onResponse-gated). The
+    // derived context's own _onResponseCallbacks holds only capture match-
+    // middleware callbacks (HTTP middleware never runs for a capture), so firing
+    // them is safe. Best-effort: a throwing callback must not fail the capture.
+    const responseCallbacks = reqCtx._onResponseCallbacks;
+    if (responseCallbacks && responseCallbacks.length > 0) {
+      const synthetic = new Response(null, { status: 200 });
+      for (const cb of responseCallbacks) {
+        try {
+          cb(synthetic);
+        } catch {
+          // A capture-time cache write that throws is degradation, not failure.
+        }
+      }
+    }
+
+    // Drain the capture data snapshot from the recording store on the derived
+    // context. Await the deferred cache writes first so a MISS-at-capture value
+    // (setItem/set scheduled under waitUntil, including the segment write just
+    // fired) is pinned, not just read-hits. When no recording store is installed
+    // (unit tests that call this directly), there is simply no snapshot.
+    const recording = getRecordingStore(reqCtx._cacheStore);
+    let snapshot: ShellSnapshotRecord[] | undefined;
+    if (recording) {
+      const settleStart = performance.now();
+      await recording.settleWrites(SHELL_SNAPSHOT_WRITE_SETTLE_MS);
+      if (stats) {
+        stats.writeSettleMs = Math.round(performance.now() - settleStart);
+      }
+      snapshot = recording.drainSnapshot();
+    }
+
+    // Pin the bake-lane loader containers (loader family). Settled containers
+    // are promise-elided (a still-pending nested promise is a hole marker, not
+    // shell material) and Flight-serialized; a REJECTED container refuses the
+    // capture — per-loader error UI must never bake into the shared shell. A
+    // container still pending here either pinned the tree (the trivial-prelude
+    // gate above already returned no-shell) or postponed under an ANCESTOR
+    // boundary (it is a hole; omitting the record keeps it live).
+    const loaderRecords = reqCtx._shellCaptureLoaderRecords;
+    // Set once a bake-lane loader settles with real (non-hole) material: its data
+    // is frozen into the shell prelude regardless of whether snapshot
+    // serialization succeeds. Drives the untagged-bake dev warning below.
+    let bakedLoaderMaterial = false;
+    if (loaderRecords && loaderRecords.size > 0) {
+      // The codec import is deferred past the elide probes: a rejected record
+      // refuses and a never-settled record is omitted WITHOUT touching Flight
+      // (also keeps the virtual @vitejs/plugin-rsc import out of unit configs).
+      let serializeContainer:
+        | typeof import("../cache/segment-codec.js").serializeResult
+        | undefined;
+      for (const [segmentKey, containerPromise] of loaderRecords) {
+        const elided = await elideLoaderContainer(containerPromise);
+        if (elided.state === "rejected") {
+          warnCaptureRefusedOnce(
+            capture.key,
+            `the loader for segment "${segmentKey}" rejected during capture; its error UI must not bake into the shared shell. ` +
+              "Fix the loader, or give its entry a loading() boundary so it stays on the live lane.",
+          );
+          return "refused";
+        }
+        // The container itself never settled: it is a hole (under an ancestor
+        // boundary) or the trivial-prelude gate already fired. Omit — no pin.
+        if (isLoaderHoleMarker(elided.value)) continue;
+        // Past the hole check: this container settled with real material that
+        // bakes into the shell prelude (independent of the snapshot pin below).
+        bakedLoaderMaterial = true;
+        try {
+          // serializeResult (not rscSerialize): null is a valid container and
+          // must round-trip; serializeResult preserves it through Flight.
+          serializeContainer ??= (await import("../cache/segment-codec.js"))
+            .serializeResult;
+          const serialized = await serializeContainer(elided.value);
+          if (serialized !== null) {
+            (snapshot ??= []).push({
+              family: "loader",
+              key: segmentKey,
+              value: { value: serialized },
+            });
+          }
+        } catch {
+          // Non-serializable container: leave it unpinned (it drifts on a HIT,
+          // the pre-snapshot behavior) rather than failing the capture.
+        }
+      }
+    }
+
+    // Snapshot size guard (issue #651): the snapshot duplicates every pinned
+    // cache value inside the shell entry, so a page over a large cache()
+    // segment can push the stored envelope toward store value limits (KV caps
+    // a value at 25 MiB) with no signal — the kv.put rejects deep inside
+    // waitUntil. Measure the serialized snapshot (UTF-8 bytes of the JSON that
+    // rides in the envelope) AFTER the loader family is appended, and over the
+    // cap store the shell WITHOUT it: pinned reads then fall back to the live
+    // store on a HIT (documented drift — hydration repairs a mismatch
+    // client-side, the pre-snapshot behavior), which beats losing the whole
+    // entry to a store-side write rejection. Reported once per key.
+    if (snapshot && snapshot.length > 0) {
+      const snapshotBytes = SNAPSHOT_BYTE_ENCODER.encode(
+        JSON.stringify(snapshot),
+      ).length;
+      if (stats) stats.snapshotBytes = snapshotBytes;
+      const cap = capture.maxSnapshotBytes ?? DEFAULT_PPR_MAX_SNAPSHOT_BYTES;
+      if (snapshotBytes > cap) {
+        warnSnapshotOverCapOnce(capture.key, snapshotBytes, cap);
+        snapshot = undefined;
+        if (stats) stats.snapshotSkipped = true;
+      }
+    }
+
+    // Shell tags snapshot at the WRITE BARRIER, not at stream construction: by
+    // here the capture has quiesced and the deferred cache writes were awaited, so
+    // tags recorded AFTER an await in async shell content (and by async
+    // cache()/"use cache" reads propagating through recordRequestTags) are
+    // included — issue #676. Loaders are masked, so loader cache tags — which
+    // belong to the holes, not the shell — never execute during capture and cannot
+    // contribute. Union with the route's static ppr.tags (capture.tags); the
+    // collected set is authoritative, the option only adds what the render cannot
+    // know.
+    const collected = [...reqCtx._requestTags];
+    const union = new Set<string>([...(capture.tags ?? []), ...collected]);
+    const shellTags = union.size > 0 ? [...union] : undefined;
+
+    // Untagged-bake diagnostic: a bake-lane loader froze mutable data into the
+    // shell but nothing tags the entry, so it is un-invalidatable except by TTL —
+    // a read-your-own-writes gap on the document channel (an action refresh skips
+    // the server shell; updateTag cannot drop untagged data). Coarse per-shell-key
+    // signal; dev-only and once-per-key so it never spams production.
+    if (isDevMode() && bakedLoaderMaterial && shellTags === undefined) {
+      warnUntaggedShellBakeOnce(capture.key);
+    }
+
+    const store = capture.store ?? reqCtx._cacheStore;
+    if (store?.putShell) {
+      try {
+        const entry: ShellCacheEntry = {
+          // slice() copies just this view's bytes into a fresh ArrayBuffer, so a
+          // prelude that is a subarray of a larger backing buffer encodes only its
+          // own region — bufferToBase64 reads the whole ArrayBuffer it is handed.
+          prelude: bufferToBase64(result.prelude.slice().buffer as ArrayBuffer),
+          postponed: result.postponed,
+          reactVersion: React.version,
+          buildVersion: capture.buildVersion,
+          // The theme this capture's payload was built with (buildFullPayload
+          // reads reqCtx.theme off the derived context). The serve tail replays
+          // it so the resume tree matches the frozen prelude — see
+          // ShellCacheEntry.initialTheme.
+          initialTheme: reqCtx.theme,
+          snapshot,
+          // Handler-layer liveness folded at the barrier: nested thenables in
+          // handler-scoped pushes, handler pushes still pending (liveness
+          // unknowable), or a handler-invoked loader execution — any of them
+          // refuses the FAST PATH, not the capture. See
+          // _shellCaptureHandleLiveness.
+          handlerLiveHoles: handlerLayerIsLive(
+            reqCtx._shellCaptureHandleLiveness,
+          ),
+          createdAt: Date.now(),
+        };
+        await store.putShell(
+          capture.key,
+          entry,
+          capture.ttl,
+          capture.swr,
+          shellTags,
+        );
+      } catch (error) {
+        // Best-effort: a failed put must never throw out of the background task.
+        reportCacheError(
+          error,
+          "cache-write",
+          "[ShellCache] capture put",
+          reqCtx,
+        );
+      }
+    }
+    // A shell was captured (the store I/O may have failed, but that is reported,
+    // not retried) — so this attempt is `stored` and the caller does not retry.
+    return "stored";
+  } finally {
+    // Stop the hop loop for the pathological never-quiets path (quiesce never
+    // fired, capture returned via maxWaitMs). On the normal path the loop already
+    // stopped when it fired quiesce; dispose() is then a no-op.
+    gate.dispose();
+  }
+}
+
+// Exported for unit tests that drive the capture core directly, and — with the
+// cold-graph retry pieces — for producer B (prerender/build-shell-capture.ts),
+// which mirrors the runtime capture's retry-in-place with the same delay.
+export {
+  runShellCapture,
+  captureAndStoreShell,
+  delay,
+  SHELL_CAPTURE_RETRY_DELAY_MS,
+};
+
+// Exported for unit tests that pin the refused-capture backoff policy directly
+// (dev cap vs production exponential growth, stored-clears, cold-start re-probe).
+// These are the same module-level functions the schedule path uses; a test that
+// drove them through a real capture round-trip could not assert the exact window
+// arithmetic without a full cold render.
+export {
+  isCaptureBackedOff,
+  markCaptureBackoff,
+  clearCaptureBackoff,
+  REFUSED_CAPTURE_BASE_MS,
+  REFUSED_CAPTURE_MAX_MS,
+  REFUSED_CAPTURE_DEV_MAX_MS,
+};