diff --git a/open-sse/executors/antigravity.ts b/open-sse/executors/antigravity.ts
index 1623b598f7..545a73ae70 100644
--- a/open-sse/executors/antigravity.ts
+++ b/open-sse/executors/antigravity.ts
@@ -224,6 +224,7 @@ export class AntigravityExecutor extends BaseExecutor {
 
     const collect = async () => {
       const chunks: string[] = [];
+      let timedOut = false;
       const timeout = AbortSignal.timeout(SSE_COLLECT_TIMEOUT_MS);
       try {
         // eslint-disable-next-line no-constant-condition
@@ -239,7 +240,9 @@ export class AntigravityExecutor extends BaseExecutor {
           chunks.push(decoder.decode(value, { stream: true }));
         }
       } catch (err) {
-        log?.warn?.("SSE_COLLECT", `Error collecting SSE stream: ${err?.message || err}`);
+        const msg = err?.message || String(err);
+        timedOut = msg.includes("timed out");
+        log?.warn?.("SSE_COLLECT", `Error collecting SSE stream: ${msg}`);
         // Fall through — return whatever was collected so far
       }
       const rawSSE = chunks.join("");
@@ -289,15 +292,16 @@ export class AntigravityExecutor extends BaseExecutor {
           {
             index: 0,
             message: { role: "assistant", content: textContent },
-            finish_reason: finishReason,
+            finish_reason: timedOut ? "length" : finishReason,
           },
         ],
         ...(usage && { usage }),
       };
 
+      const syntheticStatus = timedOut ? 504 : response.status;
       const syntheticResponse = new Response(JSON.stringify(result), {
-        status: response.status,
-        statusText: response.statusText,
+        status: syntheticStatus,
+        statusText: timedOut ? "Gateway Timeout" : response.statusText,
         headers: [["Content-Type", "application/json"]],
       });
 
diff --git a/open-sse/handlers/chatCore.ts b/open-sse/handlers/chatCore.ts
index c9612edd85..8b7176a91a 100644
--- a/open-sse/handlers/chatCore.ts
+++ b/open-sse/handlers/chatCore.ts
@@ -1217,7 +1217,7 @@ export async function handleChatCore({
           const isPassthrough = provider && getPassthroughProviders().has(provider);
           if (isPassthrough) {
             const { lockModel } = await import("../services/accountFallback.ts");
-            const cooldown = retryAfterMs || 60_000;
+            const cooldown = retryAfterMs || 120_000; // 2 min default, same as COOLDOWN_MS.rateLimit
             lockModel(provider, connectionId, model, "rate_limited", cooldown);
             console.warn(
               `[provider] Node ${connectionId} model-only rate limited (${statusCode}) for ${model} - ${Math.ceil(cooldown / 1000)}s (connection stays active)`
diff --git a/src/sse/services/auth.ts b/src/sse/services/auth.ts
index 7b24eaf19a..e465a7fd49 100644
--- a/src/sse/services/auth.ts
+++ b/src/sse/services/auth.ts
@@ -809,7 +809,7 @@ export async function markAccountUnavailable(
     // may still have quota available. Use lockModel() instead of connection-wide
     // rateLimitedUntil, same pattern as the 404 model-only lockout above.
     if (isPassthroughProvider && status === 429 && provider && model) {
-      const modelCooldown = cooldownMs || COOLDOWN_MS.rateLimited;
+      const modelCooldown = cooldownMs || COOLDOWN_MS.rateLimit;
       lockModel(provider, connectionId, model, reason || "rate_limited", modelCooldown);
       log.info(
         "AUTH",