test(e2e): un user physique par batterie — 20 min et 286s de synchro tombent à 3,6 min et 30s

La suite se ralentissait elle-même, de façon monotone. Chaque batterie crée ~11
identités virtuelles FRAÎCHES (`@alice-…`, `@owner-…`, `@recon-…`), chacune avec
ses trois documents de scope et son inbox, et toutes atterrissaient dans le MÊME
user physique — un wallet créé le 10 juillet et réutilisé depuis, que rien ne
nettoyait. Or une resynchronisation à froid est O(taille du user physique), ce
que la doc de cette bibliothèque énonce elle-même. D'où 250s il y a une semaine,
286s avant-hier, et une batterie qui a fini par dépasser les 20 minutes.

Les identités fraîches ne sont pas la faute : ce sont elles qui rendent une
batterie reproductible, une inbox stable accumulant sinon les dépôts des runs
précédents. La faute était de conserver le user physique qui les héberge.

Mesuré : 42/42 en **3,6 min** au lieu de 20+, synchro à froid la plus lente à
**30s** au lieu de 286s.

Le profil reste persistant À L'INTÉRIEUR d'une batterie — CONTRACT 1 et 2
testent précisément cela (reconnexion fidèle sur le même profil, absence de fork
de compte au travers).

Deux garde-fous pour que la prochaine dérive se voie :

- **Le budget appartient au runner**, qui échoue en nommant la cause probable.
  Un `timeout` posé autour de la commande tuait le navigateur, et la suite
  rapportait « Target page, context or browser has been closed » — un message
  qui se lit comme un défaut applicatif, et que j'ai diagnostiqué deux fois de
  travers avant de comparer les durées.
- **La synchro à froid remonte dans le résumé.** C'est le nombre qui a dérivé
  pendant un mois sans que personne le regarde, parce qu'il n'apparaissait qu'au
  détour de la ligne de détail d'une étape.

packages/client/e2e/broker.ts

@@ -72,15 +72,38 @@ export function serveHarness(): Promise<{ url: string; close: () => void }> {
 
 /**
  * Create the dedicated lib wallet once (headless UI flow on nextgraph.eu),
- * persisted in PROFILE_DIR. Mirrors Festipod's ensureAuth but with the lib's own
- * wallet name + profile. Idempotent via the ready marker.
+ * persisted in PROFILE_DIR for the duration of the batch.
+ *
+ * ── One PHYSICAL user per batch, not one forever ──────────────────────────
+ * This used to reuse a single wallet across every run, guarded by a ready marker. That
+ * made the suite slow itself down, monotonically: each batch mints ~11 FRESH virtual
+ * identities (`@alice-…`, `@owner-…`, `@recon-…`), each with three scope documents and
+ * an inbox, and they all land in the SAME physical user. Nothing ever removed them. A
+ * cold resynchronisation is O(the physical user's size) — which this library's own docs
+ * state — so the wallet created on 2026-07-10 had grown enough to take 286s on a single
+ * sync step, against 250s a week earlier, and the drift was invisible because no one
+ * measured it.
+ *
+ * The fresh identities are not the mistake — they are what makes a batch reproducible
+ * (a stable inbox accumulates its past runs' deposits otherwise). The mistake was
+ * keeping the physical user that holds them. So: a new one per batch, which also makes
+ * the cold-sync duration comparable from one run to the next instead of being a number
+ * that only ever grows.
+ *
+ * What this does NOT change: the profile stays persistent WITHIN a batch, because
+ * CONTRACT 1 and 2 test exactly that (a faithful reconnect over the same profile, and
+ * the absence of an account fork across it).
  */
 export async function ensureWallet(): Promise<void> {
   if (fs.existsSync(WALLET_READY_MARKER)) {
-    console.log("[e2e] dedicated lib wallet present — skipping creation");
-    return;
+    const age = Date.now() - fs.statSync(WALLET_READY_MARKER).mtimeMs;
+    console.log(
+      `[e2e] discarding the previous batch's wallet (${Math.round(age / 60000)} min old) — ` +
+        "a physical user is per-batch, see ensureWallet",
+    );
+    fs.rmSync(PROFILE_DIR, { recursive: true, force: true });
   }
-  console.log("[e2e] creating dedicated lib wallet on nextgraph.eu...");
+  console.log("[e2e] creating this batch's wallet on nextgraph.eu...");
   fs.mkdirSync(PROFILE_DIR, { recursive: true });
   const ctx = await chromium.launchPersistentContext(PROFILE_DIR, {
     headless: true,

packages/client/e2e/run.ts

@@ -91,6 +91,37 @@ async function faithfulReconnect(
   return { page: p, frame };
 }
 
+/**
+ * The batch's own budget, and the measurement that explains an overrun.
+ *
+ * Not a `timeout` wrapped around the command from outside: when that fired it killed the
+ * browser, and the suite reported `Target page, context or browser has been closed` —
+ * which reads as an application bug and was twice diagnosed as one. A budget belongs to
+ * the thing that knows what it is spending it on, and it must say so when it runs out.
+ */
+const BATCH_BUDGET_MS = 45 * 60 * 1000;
+const batchStart = Date.now();
+/**
+ * The slowest cold resynchronisation of the batch — the number that drifted from 250s to
+ * 286s over a month without anyone looking, because it only ever appeared inside one
+ * step's detail line. It is the health indicator of the physical user, so it is reported
+ * with the summary.
+ */
+let coldSyncMs = 0;
+
+/** Fail with the cause named, rather than letting a killed browser look like a defect. */
+function assertWithinBudget(): void {
+  const spent = Date.now() - batchStart;
+  if (spent > BATCH_BUDGET_MS) {
+    throw new Error(
+      `[e2e] batch budget exceeded (${Math.round(spent / 60000)} min > ` +
+        `${BATCH_BUDGET_MS / 60000} min). This is almost always the physical user having ` +
+        "grown: a cold resync is O(its size). Check the cold-sync figure printed above — " +
+        "it should be stable from batch to batch now that each gets a fresh wallet.",
+    );
+  }
+}
+
 async function main(): Promise<void> {
   console.log("[e2e] building SDK page bundle...");
   buildBundle();
@@ -522,7 +553,8 @@ async function main(): Promise<void> {
             `[SYNC] reconnect re-reads its OWN persisted ${scope} marker (cold-read)`,
             found,
             found
-              ? `synced in ${syncMs}ms (reconnect-login ${loginMs}ms) — ${lastDetail}`
+              ? ((coldSyncMs = Math.max(coldSyncMs, syncMs)),
+                 `synced in ${syncMs}ms (reconnect-login ${loginMs}ms) — ${lastDetail}`)
               : `NEVER synced within 120s (reconnect-login ${loginMs}ms) — ${lastDetail}`,
           );
         }
@@ -744,7 +776,14 @@ async function main(): Promise<void> {
   // ── Summary ───────────────────────────────────────────────────────────────
   const passed = results.filter((r) => r.ok).length;
   const failed = results.length - passed;
-  console.log(`\n══ SDK e2e summary: ${passed} passed, ${failed} failed, ${results.length} total ══`);
+  const batchMin = ((Date.now() - batchStart) / 60000).toFixed(1);
+  console.log(
+    `\n══ SDK e2e summary: ${passed} passed, ${failed} failed, ${results.length} total ` +
+      `— batch ${batchMin} min, slowest cold sync ${Math.round(coldSyncMs / 1000)}s ══`,
+  );
+  // A fresh wallet per batch is what should keep the cold sync flat; if it climbs from
+  // one batch to the next, the per-batch wallet is not being discarded.
+  assertWithinBudget();
   if (failed > 0) {
     console.log("Failures:");
     for (const r of results.filter((x) => !x.ok)) console.log(`  - ${r.name}: ${r.detail ?? ""}`);