Share optional MCP startup grace across connection sets (#36011)

## Why

Each connection set could restart the one-second startup grace for the same
cacheable optional MCP server, repeatedly delaying catalog capture while that
server was still pending.

## What changed

- Store the optional startup deadline in the shared tool catalog cache entry.
- Reuse that deadline until a catalog is published, and allow a fresh grace
  period after the cached catalog expires.
- Preserve per-connection-set startup grace for servers that disable catalog
  caching.

## Testing

Added a paused-time test covering shared deadlines, immediate cached-tool reuse,
cache expiration, and disabled caching.

GitOrigin-RevId: 3494969ee7002c70ce9e4ffd370e46a22d38cc11

codex-rs/codex-mcp/src/connection_manager/tool_catalog.rs

@@ -173,12 +173,16 @@ impl McpConnectionSet {
                     return;
                 }
                 if !must_wait_for_startup {
-                    if tokio::time::timeout_at(
-                        optional_startup_deadline,
-                        view.connection.client.client(),
-                    )
-                    .await
-                    .is_err()
+                    let startup_deadline = view
+                        .connection
+                        .client
+                        .tool_catalog_cache_context
+                        .as_ref()
+                        .map(|cache| cache.optional_startup_deadline(optional_startup_deadline))
+                        .unwrap_or(optional_startup_deadline);
+                    if tokio::time::timeout_at(startup_deadline, view.connection.client.client())
+                        .await
+                        .is_err()
                     {
                         trace!(server_name = %server_name, "omitting pending optional MCP server");
                     }

codex-rs/codex-mcp/src/connection_manager_tests.rs

@@ -1873,6 +1873,121 @@ async fn capture_binding_skips_pending_optional_servers_after_one_shared_startup
     assert!(binding.is_err(), "explicitly requested servers must wait");
 }
 
+#[tokio::test(start_paused = true)]
+async fn capture_binding_shares_optional_startup_grace_across_connection_sets() {
+    let approval_policy = Constrained::allow_any(AskForApproval::OnRequest);
+    let permission_profile = Constrained::allow_any(PermissionProfile::default());
+    let cache = McpToolCatalogCache::default();
+    let runtime_context = McpRuntimeContext::new(
+        Arc::new(environment_manager_without_environments()),
+        std::env::temp_dir(),
+    );
+    let server_config: McpServerConfig =
+        serde_json::from_value(serde_json::json!({ "command": "pending-mcp" }))
+            .expect("pending MCP server configuration");
+    let cache_context = cache
+        .context(
+            "pending",
+            &server_config,
+            &runtime_context,
+            /*resolved_environment*/ None,
+            &ElicitationCapability::default(),
+            /*supports_openai_form_elicitation*/ false,
+        )
+        .expect("shared pending MCP catalog");
+
+    let create_connection_set = || {
+        let mut manager = McpConnectionSet::new_uninitialized(
+            &approval_policy,
+            &permission_profile,
+            /*prefix_mcp_tool_names*/ true,
+        );
+        manager.insert_test_client(
+            "pending",
+            AsyncManagedClient {
+                client: futures::future::pending::<Result<ManagedClient, StartupOutcomeError>>()
+                    .boxed()
+                    .shared(),
+                is_codex_apps_mcp_server: false,
+                cached_server_info: None,
+                codex_apps_tools_cache_context: None,
+                tool_catalog_cache_context: Some(cache_context.clone()),
+                startup_complete: Arc::new(AtomicBool::new(false)),
+                startup_reconnect: None,
+                cancel_token: CancellationToken::new(),
+            },
+        );
+        Arc::new(manager)
+    };
+
+    let first_started = tokio::time::Instant::now();
+    let first = tokio::time::timeout(
+        Duration::from_millis(1500),
+        capture_binding(&create_connection_set()),
+    )
+    .await
+    .expect("the first thread should receive the optional startup grace");
+    assert!(first.tools().is_empty());
+    assert_eq!(first_started.elapsed(), Duration::from_secs(1));
+
+    let second = tokio::time::timeout(
+        Duration::from_millis(1),
+        capture_binding(&create_connection_set()),
+    )
+    .await
+    .expect("the next thread must not restart the same server's startup grace");
+    assert!(second.tools().is_empty());
+
+    cache_context.publish_if_newest(
+        cache_context.begin_fetch(),
+        &[create_test_tool("pending", "cached_tool")],
+    );
+    let deadline_after_publication = tokio::time::Instant::now() + Duration::from_secs(1);
+    assert_eq!(
+        cache_context.optional_startup_deadline(deadline_after_publication),
+        deadline_after_publication,
+        "publishing a catalog must not install a stale startup deadline"
+    );
+    let cached = tokio::time::timeout(
+        Duration::from_millis(1),
+        capture_binding(&create_connection_set()),
+    )
+    .await
+    .expect("cached tools should be immediately available to later threads");
+    assert_eq!(
+        cached
+            .tools()
+            .iter()
+            .map(|tool| tool.callable_name.as_str())
+            .collect::<Vec<_>>(),
+        vec!["cached_tool"]
+    );
+
+    tokio::time::advance(Duration::from_secs(30 * 60 + 1)).await;
+    assert!(
+        tokio::time::timeout(
+            Duration::from_millis(1),
+            capture_binding(&create_connection_set()),
+        )
+        .await
+        .is_err(),
+        "an expired catalog should receive a fresh startup grace"
+    );
+
+    cache_context.disable();
+    for _ in 0..2 {
+        let started = tokio::time::Instant::now();
+        let binding = tokio::time::timeout(
+            Duration::from_millis(1500),
+            capture_binding(&create_connection_set()),
+        )
+        .await
+        .expect("non-cacheable servers should keep their per-thread startup grace");
+        assert!(binding.tools().is_empty());
+        assert_eq!(started.elapsed(), Duration::from_secs(1));
+    }
+}
+
 #[tokio::test]
 async fn capture_binding_resolves_concurrently_and_rechecks_cached_clients() {
     let codex_home = tempdir().expect("tempdir");

codex-rs/codex-mcp/src/tool_catalog_cache.rs

@@ -51,6 +51,7 @@ struct ToolCatalogCacheEntry {
 #[derive(Default)]
 struct ToolCatalogCacheState {
     snapshot: Option<ToolCatalogSnapshot>,
+    optional_startup_deadline: Option<Instant>,
     last_accepted_generation: u64,
     disabled_by_server: bool,
 }
@@ -108,6 +109,21 @@ impl McpToolCatalogCacheContext {
         self.current_tools().is_some()
     }
 
+    pub(crate) fn optional_startup_deadline(&self, default_deadline: Instant) -> Instant {
+        let mut state = lock_unpoisoned(&self.entry.state);
+        if state.disabled_by_server
+            || state
+                .snapshot
+                .as_ref()
+                .is_some_and(|snapshot| snapshot.published_at.elapsed() <= TOOL_CATALOG_CACHE_TTL)
+        {
+            return default_deadline;
+        }
+        *state
+            .optional_startup_deadline
+            .get_or_insert(default_deadline)
+    }
+
     pub(crate) fn current_tools(&self) -> Option<Vec<ToolInfo>> {
         lock_unpoisoned(&self.entry.state)
             .snapshot
@@ -147,6 +163,7 @@ impl McpToolCatalogCacheContext {
             tool.tool.annotations = None;
         }
         state.last_accepted_generation = ticket.generation;
+        state.optional_startup_deadline = None;
         state.snapshot = Some(ToolCatalogSnapshot {
             tools,
             published_at: Instant::now(),