diff --git a/packages/app-core/src/state/startup-phase-hydrate-nonblocking.test.ts b/packages/app-core/src/state/startup-phase-hydrate-nonblocking.test.ts
new file mode 100644
index 0000000000..a5395a21d8
--- /dev/null
+++ b/packages/app-core/src/state/startup-phase-hydrate-nonblocking.test.ts
@@ -0,0 +1,133 @@
+/**
+ * Regression: runHydrating() must reach HYDRATION_COMPLETE without blocking on
+ * slow/hanging non-critical work.
+ *
+ * LANE F2 sol-f5-firstload (2026-07-22): post-login "takes forever" was caused
+ * by runHydrating() serially AWAITING work the landing view never needs:
+ *   - getWalletAddresses()  (measured 1.3-12s on cloud containers)
+ *   - a VRM + gaussian-splat world prefetch race capped at 15s
+ *   - fetchAutonomyReplay()
+ * These now run in the background. This test proves the dashboard becomes
+ * interactive (HYDRATION_COMPLETE dispatched) even when every one of those
+ * hangs indefinitely.
+ */
+import { describe, expect, it, vi } from "vitest";
+
+// ── Mocks for module singletons runHydrating touches ──────────────────
+const hangForever = () => new Promise<never>(() => {});
+
+vi.mock("../api", () => ({
+  client: {
+    // The hangers — must NOT block hydration:
+    getWalletAddresses: vi.fn(() => hangForever()),
+    // Fast, awaited config reads (parallelized):
+    getConfig: vi.fn(async () => ({ ui: {} })),
+    getStreamSettings: vi.fn(async () => ({ settings: {} })),
+    hasCustomVrm: vi.fn(async () => false),
+    hasCustomBackground: vi.fn(async () => false),
+  },
+}));
+
+vi.mock("../components/avatar/VrmEngine", () => ({
+  // VRM prefetch hangs forever — must NOT block hydration.
+  prefetchVrmToCache: vi.fn(() => hangForever()),
+}));
+
+vi.mock("./vrm", () => ({
+  getVrmUrl: (i: number) => `vrm://${i}`,
+  getVrmCount: () => 2,
+  VRM_COUNT: 2,
+}));
+
+vi.mock("./persistence", () => ({ loadUiTheme: () => "dark" }));
+
+vi.mock("../utils", () => ({
+  resolveApiUrl: (p: string) => p,
+  resolveAppAssetUrl: (p: string) => `asset://${p}`,
+}));
+
+// Keep the rest of the module's imports cheap/no-op.
+vi.mock("./internal", () => ({
+  loadAvatarIndex: () => 1,
+  normalizeAvatarIndex: (n: number) => n,
+}));
+vi.mock("./shell-routing", () => ({
+  shouldStartAtCharacterSelectOnLaunch: () => false,
+}));
+vi.mock("../navigation", () => ({
+  COMPANION_ENABLED: true,
+  tabFromPath: () => null,
+  isRouteRootPath: () => true,
+}));
+
+// world prefetch uses global fetch — make it hang too.
+vi.stubGlobal("fetch", vi.fn(() => hangForever()));
+
+import { runHydrating } from "./startup-phase-hydrate";
+
+function makeDeps() {
+  const noop = () => {};
+  const anoop = async () => {};
+  return {
+    setStartupError: vi.fn(),
+    setOnboardingLoading: vi.fn(),
+    hydrateInitialConversationState: vi.fn(async () => null),
+    requestGreetingWhenRunningRef: { current: async () => {} },
+    loadWorkbench: vi.fn(anoop),
+    loadPlugins: vi.fn(anoop),
+    loadSkills: vi.fn(anoop),
+    loadCharacter: vi.fn(anoop),
+    loadWalletConfig: vi.fn(anoop),
+    loadInventory: vi.fn(anoop),
+    loadUpdateStatus: vi.fn(anoop),
+    checkExtensionStatus: vi.fn(anoop),
+    pollCloudCredits: vi.fn(noop),
+    fetchAutonomyReplay: vi.fn(() => hangForever()),
+    setSelectedVrmIndex: vi.fn(),
+    setCustomVrmUrl: vi.fn(),
+    setCustomBackgroundUrl: vi.fn(),
+    setWalletAddresses: vi.fn(),
+    setTab: vi.fn(),
+    setTabRaw: vi.fn(),
+    onboardingCompletionCommittedRef: { current: false },
+    initialTabSetRef: { current: false },
+    onboardingMode: "cloud" as unknown as never,
+  };
+}
+
+describe("runHydrating — non-blocking first-load (F2)", () => {
+  it("dispatches HYDRATION_COMPLETE even when wallet, VRM prefetch, world prefetch, and autonomy replay all hang", async () => {
+    const deps = makeDeps();
+    const dispatch = vi.fn();
+    const cancelled = { current: false };
+
+    // Must resolve quickly. If any hang is awaited, this rejects on timeout.
+    await Promise.race([
+      // biome-ignore lint/suspicious/noExplicitAny: test deps shape
+      runHydrating(deps as any, dispatch, cancelled),
+      new Promise((_r, reject) =>
+        setTimeout(
+          () => reject(new Error("runHydrating blocked on non-critical work")),
+          3000,
+        ),
+      ),
+    ]);
+
+    expect(dispatch).toHaveBeenCalledWith({ type: "HYDRATION_COMPLETE" });
+  });
+
+  it("does not block hydration on the slow wallet fetch, but still kicks it off", async () => {
+    const deps = makeDeps();
+    const dispatch = vi.fn();
+    await runHydrating(
+      // biome-ignore lint/suspicious/noExplicitAny: test deps shape
+      deps as any,
+      dispatch,
+      { current: false },
+    );
+    // Wallet setter never resolves (fetch hangs) so it should not have been
+    // called by the time hydration completes — proving it was deferred.
+    expect(deps.setWalletAddresses).not.toHaveBeenCalled();
+    expect(dispatch).toHaveBeenCalledWith({ type: "HYDRATION_COMPLETE" });
+  });
+});
diff --git a/packages/app-core/src/state/startup-phase-hydrate.ts b/packages/app-core/src/state/startup-phase-hydrate.ts
index 2b7553cea2..0e9c79265d 100644
--- a/packages/app-core/src/state/startup-phase-hydrate.ts
+++ b/packages/app-core/src/state/startup-phase-hydrate.ts
@@ -153,21 +153,36 @@ export async function runHydrating(
   void deps.loadPlugins();
   void deps.loadCharacter();
 
-  // Wallet addresses
-  try {
-    deps.setWalletAddresses(await client.getWalletAddresses());
-  } catch (e) {
-    warn("wallet addresses", e);
-  }
+  // ── Wallet addresses — DEFERRED, not awaited ───────────────────────
+  // Wallet addresses are only consumed by the Eliza Cloud dashboard and
+  // the wallets/inventory tab — never by the landing (chat/companion)
+  // view. getWalletAddresses() has been measured at 1.3–12 s on cloud
+  // containers (steward round-trip), so awaiting it here needlessly
+  // stalled time-to-interactive by multiple seconds. Fire-and-forget:
+  // the state populates when it resolves; the dashboard reads it lazily.
+  void (async () => {
+    try {
+      deps.setWalletAddresses(await client.getWalletAddresses());
+    } catch (e) {
+      warn("wallet addresses", e);
+    }
+  })();
 
-  // Avatar / VRM selection — resolve from server config, then stream
-  // settings, then localStorage.  Cloud containers that skip onboarding
-  // have their character defaults written server-side, so we must read
-  // the config to pick up the correct avatarIndex.
+  // ── Avatar / VRM selection ─────────────────────────────────────────
+  // Resolve from server config, then stream settings, then localStorage.
+  // Cloud containers that skip onboarding have their character defaults
+  // written server-side, so we must read the config to pick up the
+  // correct avatarIndex. config + streamSettings are INDEPENDENT reads —
+  // run them in parallel instead of serially (was ~2× the round-trip).
   let resolvedIdx = loadAvatarIndex();
-  try {
-    const cfg = await client.getConfig();
-    const cfgUi = cfg?.ui as Record<string, unknown> | undefined;
+  const [cfgResult, streamResult] = await Promise.allSettled([
+    client.getConfig(),
+    typeof client.getStreamSettings === "function"
+      ? client.getStreamSettings()
+      : Promise.resolve(null),
+  ]);
+  if (cfgResult.status === "fulfilled") {
+    const cfgUi = cfgResult.value?.ui as Record<string, unknown> | undefined;
     const cfgAvatarIdx = cfgUi?.avatarIndex;
     if (typeof cfgAvatarIdx === "number" && Number.isFinite(cfgAvatarIdx)) {
       const normalized = normalizeAvatarIndex(cfgAvatarIdx);
@@ -176,26 +191,29 @@ export async function runHydrating(
         deps.setSelectedVrmIndex(resolvedIdx);
       }
     }
-  } catch (e) {
-    warn("config avatar index", e);
+  } else {
+    warn("config avatar index", cfgResult.reason);
   }
-  try {
-    if (typeof client.getStreamSettings === "function") {
-      const stream = await client.getStreamSettings();
-      const si = stream.settings?.avatarIndex;
-      if (typeof si === "number" && Number.isFinite(si)) {
-        resolvedIdx = normalizeAvatarIndex(si);
-        deps.setSelectedVrmIndex(resolvedIdx);
-      }
+  // stream settings win over config when present
+  if (streamResult.status === "fulfilled" && streamResult.value) {
+    const si = streamResult.value.settings?.avatarIndex;
+    if (typeof si === "number" && Number.isFinite(si)) {
+      resolvedIdx = normalizeAvatarIndex(si);
+      deps.setSelectedVrmIndex(resolvedIdx);
     }
-  } catch (e) {
-    warn("stream settings avatar", e);
+  } else if (streamResult.status === "rejected") {
+    warn("stream settings avatar", streamResult.reason);
   }
   if (resolvedIdx === 0) {
-    if (await client.hasCustomVrm())
+    // custom vrm + custom background probes are independent — parallelize.
+    const [hasVrm, hasBg] = await Promise.all([
+      client.hasCustomVrm().catch(() => false),
+      client.hasCustomBackground().catch(() => false),
+    ]);
+    if (hasVrm)
       deps.setCustomVrmUrl(resolveApiUrl(`/api/avatar/vrm?t=${Date.now()}`));
     else deps.setSelectedVrmIndex(1);
-    if (await client.hasCustomBackground())
+    if (hasBg)
       deps.setCustomBackgroundUrl(
         resolveApiUrl(`/api/avatar/background?t=${Date.now()}`),
       );
@@ -209,33 +227,26 @@ export async function runHydrating(
   // noticeable in cloud containers where the CDN round-trip is the
   // bottleneck.
   //
-  // We await the active VRM prefetch (with a 15s timeout) rather than
-  // firing and forgetting. This ensures the in-memory buffer cache is
-  // populated *before* HYDRATION_COMPLETE, so the companion scene gets
-  // an instant cache hit instead of starting a duplicate network download.
-  //
-  // Additionally, fire-and-forget prefetches for ALL other VRM assets so
-  // navigating to the customize/character page doesn't trigger a full
-  // re-download of every character model.
+  // Previously this AWAITED the active VRM prefetch (with a 15 s timeout)
+  // before dispatching HYDRATION_COMPLETE — blocking the dashboard from
+  // becoming interactive for up to 15 s while 3D assets downloaded. That
+  // is the single largest post-login stall on cold cache. We now warm the
+  // cache in the BACKGROUND (fire-and-forget) and let hydration complete
+  // immediately: the companion scene renders as soon as its assets arrive,
+  // and because loadGltfAsset joins in-flight downloads via the inflight
+  // dedup map, the scene still gets the warmed buffer with no duplicate
+  // fetch — we simply no longer make the ENTIRE dashboard wait on it.
   if (COMPANION_ENABLED) {
     const vrmIdx = resolvedIdx > 0 ? resolvedIdx : 1;
-    const vrmPrefetch = prefetchVrmToCache(getVrmUrl(vrmIdx));
+    // Active companion VRM — warm in-memory buffer cache (background).
+    void prefetchVrmToCache(getVrmUrl(vrmIdx));
+    // Gaussian-splat world — warm the browser HTTP cache (background).
     const theme = loadUiTheme();
     const worldUrl =
       theme === "dark"
         ? resolveAppAssetUrl("worlds/companion-night.spz")
         : resolveAppAssetUrl("worlds/companion-day.spz");
-    const worldPrefetch = fetch(worldUrl, { cache: "force-cache" }).catch(
-      () => {},
-    );
-    // Wait for both but cap at 15s so hydration isn't blocked forever on
-    // slow networks. Even if the timeout fires, the in-flight prefetch
-    // continues in the background and loadGltfAsset will join it via the
-    // inflight dedup map.
-    await Promise.race([
-      Promise.all([vrmPrefetch, worldPrefetch]),
-      new Promise((resolve) => setTimeout(resolve, 15_000)),
-    ]);
+    void fetch(worldUrl, { cache: "force-cache" }).catch(() => {});
 
     // Fire-and-forget: warm the cache for all other VRM assets so the
     // customize page does not need to re-download them on first visit.
@@ -248,7 +259,8 @@ export async function runHydrating(
   }
 
   void deps.pollCloudCredits();
-  await deps.fetchAutonomyReplay();
+  // Autonomy replay is not needed for first paint — don't block on it.
+  void deps.fetchAutonomyReplay();
 
   // Tab routing
   const navPath = getNavigationPathFromWindow();