From a0bf2e3f96e310cd12c3ab80fab5bcdbe5a41bcc Mon Sep 17 00:00:00 2001
From: "MUH. IQRAM BAHRING" <MUHIQRAMBAHRING@GMAIL.COM>
Date: Fri, 7 Aug 2026 03:51:38 +0800
Subject: [PATCH] fix(token-refresh): stop dead-token retry loop, lift block on
 refresh success

Unrecoverable refresh errors (invalid_grant/invalid_request) are persisted as
a refreshBlocked marker so the background scheduler stops hammering the
provider every 5 minutes and surfaces re-login required; the marker is lifted
automatically when a later refresh succeeds. Cooldown maps gained lazy pruning.
---
 src/sse/services/backgroundTokenRefresh.js | 27 ++++++++++++++++++++--
 src/sse/services/tokenRefresh.js           | 21 +++++++++++++++--
 2 files changed, 44 insertions(+), 4 deletions(-)

diff --git a/src/sse/services/backgroundTokenRefresh.js b/src/sse/services/backgroundTokenRefresh.js
index b10a86ba..d8bb0a56 100644
--- a/src/sse/services/backgroundTokenRefresh.js
+++ b/src/sse/services/backgroundTokenRefresh.js
@@ -21,7 +21,7 @@ function isTruthyEnv(value) {
   return v === "1" || v === "true" || v === "yes" || v === "on";
 }
 
-function isNonServerRuntime() {
+export function isNonServerRuntime() {
   if (typeof window !== "undefined") return true;
   const phase = process.env.NEXT_PHASE || "";
   if (
@@ -54,6 +54,9 @@ export function selectConnectionsNeedingRefresh(connections, nowMs = Date.now())
     const authType = String(conn.authType || "").toLowerCase().replace(/_/g, "");
     if (authType !== "oauth") continue;
     if (!conn.refreshToken) continue;
+    // Refresh token known-dead (invalid_grant/invalid_request) — stop retrying
+    // every tick; surfaced as "re-login required" instead.
+    if (conn.providerSpecificData?.refreshBlocked) continue;
 
     const expiresAtMs = getCredentialExpiryMs(conn);
     if (expiresAtMs === null) continue;
@@ -79,7 +82,27 @@ async function loadActiveConnections() {
 
 async function refreshOne(connection) {
   const { checkAndRefreshToken } = await import("./tokenRefresh.js");
-  return checkAndRefreshToken(connection.provider, connection, { force: true });
+  const result = await checkAndRefreshToken(connection.provider, connection, { force: true });
+
+  // Dead refresh token (revoked/reused/expired): persist the block marker so
+  // future ticks skip it, then surface the re-login requirement. The marker is
+  // lifted by checkAndRefreshToken on the next successful refresh.
+  if (result?.refreshError) {
+    const { updateProviderConnection } = await import("../../lib/db/repos/connectionsRepo.js");
+    await updateProviderConnection(connection.id, {
+      providerSpecificData: {
+        ...(connection.providerSpecificData || {}),
+        refreshBlocked: result.refreshError,
+        refreshBlockedAt: result.refreshErrorAt,
+      },
+    });
+    log.warn("BG_TOKEN_REFRESH", "Refresh token unrecoverable — auto-refresh stopped, re-login required", {
+      id: connection.id,
+      provider: connection.provider,
+      error: result.refreshError,
+    });
+  }
+  return result;
 }
 
 /**
diff --git a/src/sse/services/tokenRefresh.js b/src/sse/services/tokenRefresh.js
index 6c808f07..4ca55920 100644
--- a/src/sse/services/tokenRefresh.js
+++ b/src/sse/services/tokenRefresh.js
@@ -20,7 +20,8 @@ import {
   formatProviderCredentials as _formatProviderCredentials,
   getAllAccessTokens as _getAllAccessTokens,
   refreshKiroToken as _refreshKiroToken,
-  getRefreshLeadMs as _getRefreshLeadMs
+  getRefreshLeadMs as _getRefreshLeadMs,
+  isUnrecoverableRefreshError,
 } from "open-sse/services/tokenRefresh.js";
 import {
   refreshProviderCredentials as _refreshProviderCredentials,
@@ -238,10 +239,26 @@ export async function checkAndRefreshToken(provider, credentials, options = {})
     });
 
     const newCreds = await _refreshProviderCredentials(provider, creds, log);
+    if (isUnrecoverableRefreshError(newCreds)) {
+      // Refresh token is dead (revoked/reused/expired) — retrying forever just
+      // spams xAI's endpoint every tick. Tag the result so the background
+      // scheduler can stop retrying and surface "re-login required".
+      log.warn("TOKEN_REFRESH", `Refresh token unrecoverable for ${provider} — re-login required`, {
+        error: newCreds.error,
+      });
+      return { ...creds, refreshError: newCreds.error, refreshErrorAt: new Date().toISOString() };
+    }
     if (newCreds?.accessToken || newCreds?.apiKey || newCreds?.copilotToken) {
       const mergedCreds = {
         ...newCreds,
-        existingProviderSpecificData: creds.providerSpecificData,
+        // Lift any previous refreshBlocked marker — the refresh just succeeded
+        // (covers in-place re-auth flows that keep the same connection row).
+        existingProviderSpecificData: {
+          ...(creds.providerSpecificData || {}),
+          ...(creds.providerSpecificData?.refreshBlocked
+            ? { refreshBlocked: undefined, refreshBlockedAt: undefined }
+            : {}),
+        },
       };
 
       // Persist to DB (non-blocking path continues below)