fix(analytics): address PR review — add endpoint check and document trade-offs

Author: waleedlatif1

apps/sim/lib/analytics/profound.ts

@@ -33,7 +33,7 @@ let flushTimer: NodeJS.Timeout | null = null
  * Returns true if Profound analytics is configured.
  */
 export function isProfoundEnabled(): boolean {
-  return isHosted && Boolean(env.PROFOUND_API_KEY)
+  return isHosted && Boolean(env.PROFOUND_API_KEY) && Boolean(env.PROFOUND_ENDPOINT)
 }
 
 /**
@@ -69,6 +69,8 @@ async function flush(): Promise<void> {
       logger.error(`Profound API returned ${response.status}`)
     }
   } catch (error) {
+    // Entries are intentionally not re-queued on failure to prevent unbounded memory growth.
+    // Under a Profound outage, analytics data is lost — acceptable for non-critical telemetry.
     logger.error('Failed to flush logs to Profound', error)
   }
 }

apps/sim/proxy.ts

@@ -199,6 +199,9 @@ export async function proxy(request: NextRequest) {
 
 /**
  * Sends request data to Profound analytics (fire-and-forget) and returns the response.
+ * Note: `NextResponse.next()` always carries status 200 — it signals "continue to route handler",
+ * not the final HTTP status. Only redirects (307/308) and explicit blocks (403) have accurate codes.
+ * This matches the Vercel log drain behavior where proxy-level status reflects middleware outcome.
  */
 function track(request: NextRequest, response: NextResponse): NextResponse {
   if (isProfoundEnabled()) {