From 645709bb62dd1917376ddb1118f528a2c710a806 Mon Sep 17 00:00:00 2001
From: Codex <codex@noreply.local>
Date: Sun, 28 Jun 2026 02:24:56 +0000
Subject: [PATCH] fix: retry control page recovery hydration

---
 .../hwlab-node-web-observe-runner-source.ts   | 39 ++++++++++++-------
 1 file changed, 24 insertions(+), 15 deletions(-)

diff --git a/scripts/src/hwlab-node-web-observe-runner-source.ts b/scripts/src/hwlab-node-web-observe-runner-source.ts
index 6e2cabea..10bd74ef 100644
--- a/scripts/src/hwlab-node-web-observe-runner-source.ts
+++ b/scripts/src/hwlab-node-web-observe-runner-source.ts
@@ -1655,7 +1655,7 @@ async function controlPageLivenessSnapshot(reason, timeoutMs = 1500) {
 
 async function ensureControlPageResponsiveForCommand(reason) {
   const beforeUrl = currentPageUrl();
-  const liveness = await controlPageLivenessSnapshot(reason + "-preflight", 1500);
+  const liveness = await controlPageLivenessSnapshot(reason + "-preflight", 3000);
   if (liveness.ok) return { ok: true, recovered: false, reason, beforeUrl, afterUrl: currentPageUrl(), liveness, pageRole: "control", pageId, pageEpoch: controlPageEpoch, valuesRedacted: true };
   const target = controlPageRecoveryTarget(liveness.snapshot, beforeUrl);
   await appendJsonl(files.control, eventRecord("control-page-unresponsive-before-command", {
@@ -1668,24 +1668,32 @@ async function ensureControlPageResponsiveForCommand(reason) {
     pageEpoch: controlPageEpoch,
     valuesRedacted: true
   }));
-  await recreateControlPageForNavigation(reason + "-control-page-unresponsive", 1);
   let navigation = null;
   let hydration = null;
   let afterLiveness = null;
-  try {
-    navigation = await gotoTarget(target.targetPath);
-  } catch (error) {
-    navigation = { ok: false, targetPath: target.targetPath, error: errorSummary(error), valuesRedacted: true };
+  const attempts = [];
+  let ok = false;
+  for (let attempt = 1; attempt <= 2; attempt += 1) {
+    await recreateControlPageForNavigation(reason + "-control-page-unresponsive", attempt);
+    try {
+      navigation = await gotoTarget(target.targetPath);
+    } catch (error) {
+      navigation = { ok: false, targetPath: target.targetPath, error: errorSummary(error), valuesRedacted: true };
+    }
+    if (!navigation?.error && target.sessionId) {
+      hydration = await withHardTimeout(
+        waitForWorkbenchSessionHydrated(page, target.sessionId, { timeoutMs: 12000 }),
+        14000,
+        "control page recovery hydration exceeded 14000ms"
+      ).catch((error) => ({ ok: false, error: errorSummary(error), valuesRedacted: true }));
+    } else {
+      hydration = null;
+    }
+    afterLiveness = await controlPageLivenessSnapshot(reason + "-post-recovery", 3000);
+    ok = !navigation?.error && afterLiveness.ok === true && (!target.sessionId || hydration?.ok === true);
+    attempts.push({ attempt, ok, navigation, hydration, afterLiveness, pageId, pageEpoch: controlPageEpoch, valuesRedacted: true });
+    if (ok) break;
   }
-  if (!navigation?.error && target.sessionId) {
-    hydration = await withHardTimeout(
-      waitForWorkbenchSessionHydrated(page, target.sessionId, { timeoutMs: 10000 }),
-      12000,
-      "control page recovery hydration exceeded 12000ms"
-    ).catch((error) => ({ ok: false, error: errorSummary(error), valuesRedacted: true }));
-  }
-  afterLiveness = await controlPageLivenessSnapshot(reason + "-post-recovery", 3000);
-  const ok = !navigation?.error && afterLiveness.ok === true && (!target.sessionId || hydration?.ok === true);
   const recovery = {
     ok,
     recovered: ok,
@@ -1697,6 +1705,7 @@ async function ensureControlPageResponsiveForCommand(reason) {
     navigation,
     hydration,
     afterLiveness,
+    attempts,
     pageRole: "control",
     pageId,
     pageEpoch: controlPageEpoch,