diff --git a/Tests/OpenClawKitTests/CoreAIModelRuntimeTests.swift b/Tests/OpenClawKitTests/CoreAIModelRuntimeTests.swift
index ddcf386..f98005f 100644
--- a/Tests/OpenClawKitTests/CoreAIModelRuntimeTests.swift
+++ b/Tests/OpenClawKitTests/CoreAIModelRuntimeTests.swift
@@ -296,7 +296,7 @@ struct CoreAIModelRuntimeTests {
         #expect(texts == ["aaaaa", "bbbbb"])
     }
 
-    @Test
+    @Test(.timeLimit(.minutes(1)))
     func cancelStopsOnlyTheRunningGeneration() async throws {
         let executor = SharedCacheExecutor(limit: 40)
         let engine = CoreAILocalModelEngine(tokenizer: ScalarTokenizer(), executorFactory: { _, _ in executor })
@@ -304,9 +304,7 @@ struct CoreAIModelRuntimeTests {
         configuration.temperature = 0
         try await engine.loadModel(path: "/m", configuration: configuration)
         let running = Task { try await engine.generate(prompt: "a", systemPrompt: nil, configuration: configuration, onToken: nil) }
-        for _ in 0..<500 where await executor.runs == 0 {
-            try await Task.sleep(nanoseconds: 1_000_000)
-        }
+        try await waitUntil("first generation running") { await executor.runs > 0 }
         let queued = Task { try await engine.generate(prompt: "b", systemPrompt: nil, configuration: configuration, onToken: nil) }
         await engine.cancelGeneration(token: nil)
         await #expect(throws: CoreAIRuntimeError.cancelled) {
diff --git a/Tests/OpenClawKitTests/GatewayServerChatUICompatTests.swift b/Tests/OpenClawKitTests/GatewayServerChatUICompatTests.swift
index 6d0186c..3191832 100644
--- a/Tests/OpenClawKitTests/GatewayServerChatUICompatTests.swift
+++ b/Tests/OpenClawKitTests/GatewayServerChatUICompatTests.swift
@@ -5,7 +5,7 @@ import Testing
 
 /// The in-process server's session rows, `sessions.changed`, `session.message` and protocol-v4
 /// `chat` payloads decode through the ChatUI transport models.
-@Suite("Gateway server ChatUI compatibility")
+@Suite("Gateway server ChatUI compatibility", .timeLimit(.minutes(1)))
 struct GatewayServerChatUICompatTests {
     private func makeStack(_ name: String, turns: [String]) async throws -> (GatewayServer, EmbeddedAgentRuntime, URL) {
         let root = FileManager.default.temporaryDirectory.appendingPathComponent("\(name)-\(UUID().uuidString)")
@@ -43,11 +43,8 @@ struct GatewayServerChatUICompatTests {
         private(set) var frames: [EventFrame] = []
         func record(_ frame: EventFrame) { self.frames.append(frame) }
 
-        func wait(_ predicate: @Sendable ([EventFrame]) -> Bool) async -> [EventFrame] {
-            for _ in 0..<500 {
-                if predicate(self.frames) { return self.frames }
-                try? await Task.sleep(nanoseconds: 10_000_000)
-            }
+        func wait(_ label: String, _ predicate: @escaping @Sendable ([EventFrame]) -> Bool) async throws -> [EventFrame] {
+            try await waitUntil(label) { predicate(await self.frames) }
             return self.frames
         }
     }
@@ -101,7 +98,7 @@ struct GatewayServerChatUICompatTests {
         let send = try GatewayPayloadCodec.decode(OpenClawChatSendResponse.self, from: ack.payload)
         #expect(send.runId == "ui-run-1")
 
-        let frames = await recorder.wait { frames in
+        let frames = try await recorder.wait("final chat, both session messages and the end change") { frames in
             frames.contains { $0.event == "chat" && $0.payload?.dictionaryValue?["state"] == AnyCodable("final") }
                 && frames.filter { $0.event == "session.message" }.count >= 2
                 && frames.contains { $0.event == "sessions.changed" && $0.payload?.dictionaryValue?["phase"] == AnyCodable("end") }
diff --git a/Tests/OpenClawKitTests/MediaPipelineTests.swift b/Tests/OpenClawKitTests/MediaPipelineTests.swift
index 7e6f64d..315d30c 100644
--- a/Tests/OpenClawKitTests/MediaPipelineTests.swift
+++ b/Tests/OpenClawKitTests/MediaPipelineTests.swift
@@ -7,7 +7,7 @@ import Glibc
 import Darwin
 #endif
 
-@Suite("Media pipeline", .serialized)
+@Suite("Media pipeline", .serialized, .timeLimit(.minutes(1)))
 struct MediaPipelineTests {
     @Test
     func normalizeRejectsEmptyMimeTypeAndOversizedPayloads() async throws {
@@ -415,22 +415,26 @@ struct MediaPipelineTests {
         }
 
         let baseURL = URL(string: "http://127.0.0.1:\(port)/")!
-        try await self.waitForHTTPServer(baseURL: baseURL)
+        try await self.waitForHTTPServer(baseURL: baseURL, process: process)
         try await body(baseURL)
     }
 
-    private func waitForHTTPServer(baseURL: URL) async throws {
-        var lastError: (any Error)?
-        for _ in 0..<20 {
-            do {
-                _ = try Data(contentsOf: baseURL)
+    /// Polls until the server answers. There is no deadline: a loaded runner only delays Python's
+    /// startup, and the suite's time limit bounds a server that never answers.
+    private func waitForHTTPServer(baseURL: URL, process: Process) async throws {
+        while !Task.isCancelled {
+            if (try? Data(contentsOf: baseURL)) != nil {
                 return
-            } catch {
-                lastError = error
-                try await Task.sleep(nanoseconds: 100_000_000)
             }
+            guard process.isRunning else {
+                throw OpenClawCoreError.unavailable("Local HTTP server exited with status \(process.terminationStatus)")
+            }
+            try? await Task.sleep(nanoseconds: 100_000_000)
         }
-        throw lastError ?? OpenClawCoreError.unavailable("Local HTTP server did not start")
+        // Swift Testing drops errors thrown after a time-limit cancellation, so record which wait hung.
+        let timeout = AsyncWaitTimeoutError(label: "local HTTP server answering")
+        Issue.record(timeout)
+        throw timeout
     }
 
     private func reserveTCPPort() throws -> Int {
diff --git a/Tests/OpenClawKitTests/ModelRoutingTests.swift b/Tests/OpenClawKitTests/ModelRoutingTests.swift
index 6b22ef8..c5bb624 100644
--- a/Tests/OpenClawKitTests/ModelRoutingTests.swift
+++ b/Tests/OpenClawKitTests/ModelRoutingTests.swift
@@ -332,9 +332,11 @@ struct ModelRoutingTests {
                 StaticProvider(id: "primary", text: "primary-output"),
                 StaticProvider(id: "secondary", text: "secondary-output"),
             ],
+            // An hour-long window: the second request is inside it however slow the runner is, and a
+            // drop never waits for the window to pass.
             throttlePolicy: ModelProviderThrottlePolicy(
                 maxRequestsPerWindow: 1,
-                windowMs: 1_000,
+                windowMs: 3_600_000,
                 strategy: .drop
             ),
             diagnosticsSink: await pipeline.sink()
@@ -355,9 +357,9 @@ struct ModelRoutingTests {
         #expect(events.contains(where: { $0.name == "model.request.retry" }))
     }
 
-    @Test
+    @Test(.timeLimit(.minutes(1)))
     func routerThrottleDelayAppliesCooldownAndEmitsDiagnostics() async throws {
-        let pipeline = RuntimeDiagnosticsPipeline(eventLimit: 50)
+        let request = ModelGenerationRequest(sessionKey: "main", prompt: "hello", providerID: "primary")
         let router = ModelRouter(
             defaultProviderID: "primary",
             providers: [StaticProvider(id: "primary", text: "primary-output")],
@@ -365,22 +367,39 @@ struct ModelRoutingTests {
                 maxRequestsPerWindow: 1,
                 windowMs: 70,
                 strategy: .delay
-            ),
-            diagnosticsSink: await pipeline.sink()
+            )
         )
 
+        // The second request proceeds once the window has passed. A lower bound: a slow runner only
+        // adds to it (and may let the window pass before the second request arrives).
         let startedAt = Date()
-        _ = try await router.generate(
-            ModelGenerationRequest(sessionKey: "main", prompt: "hello", providerID: "primary")
-        )
-        _ = try await router.generate(
-            ModelGenerationRequest(sessionKey: "main", prompt: "hello", providerID: "primary")
-        )
-        let elapsed = Date().timeIntervalSince(startedAt)
+        _ = try await router.generate(request)
+        _ = try await router.generate(request)
+        #expect(Date().timeIntervalSince(startedAt) >= 0.06)
 
-        #expect(elapsed >= 0.06)
-        let events = await pipeline.recentEvents(limit: 50)
-        #expect(events.contains(where: { $0.name == "model.throttle.delay" }))
+        // With an hour-long window the second request always lands inside it and waits.
+        let pipeline = RuntimeDiagnosticsPipeline(eventLimit: 50)
+        let slowRouter = ModelRouter(
+            defaultProviderID: "primary",
+            providers: [StaticProvider(id: "primary", text: "primary-output")],
+            throttlePolicy: ModelProviderThrottlePolicy(
+                maxRequestsPerWindow: 1,
+                windowMs: 3_600_000,
+                strategy: .delay
+            ),
+            diagnosticsSink: await pipeline.sink()
+        )
+        _ = try await slowRouter.generate(request)
+        let delayed = Task { try await slowRouter.generate(request) }
+        try await waitUntil("model.throttle.delay emitted") {
+            await pipeline.recentEvents(limit: 50).contains { $0.name == "model.throttle.delay" }
+        }
+        let delay = try #require(await pipeline.recentEvents(limit: 50).first { $0.name == "model.throttle.delay" })
+        #expect((Int(delay.metadata["delayMs"] ?? "") ?? 0) > 3_000_000)
+        delayed.cancel()
+        await #expect(throws: CancellationError.self) {
+            _ = try await delayed.value
+        }
     }
 
     @Test
diff --git a/Tests/OpenClawKitTests/SystemStateReportingTests.swift b/Tests/OpenClawKitTests/SystemStateReportingTests.swift
index a2f88f9..00f681b 100644
--- a/Tests/OpenClawKitTests/SystemStateReportingTests.swift
+++ b/Tests/OpenClawKitTests/SystemStateReportingTests.swift
@@ -191,16 +191,17 @@ struct SystemStateReportingTests {
         #expect(recorder.reports.last == .volatile(.gateway, ["backoffMs": 4000]))
     }
 
-    @Test
+    @Test(.timeLimit(.minutes(1)))
     func coalescingReporterSchedulesTrailingFlush() async throws {
         let recorder = RecordingStateReporter()
-        let reporter = CoalescingSystemStateReporter(wrapping: recorder, minimumInterval: 0.05)
+        // The clock stands still, so the second update always lands inside the interval; only the
+        // trailing flush can forward it.
+        let clock = TestClock()
+        let reporter = CoalescingSystemStateReporter(wrapping: recorder, minimumInterval: 0.05, now: { clock.now })
         reporter.reportVolatileUpdate(.talk, ["level": 1])
         reporter.reportVolatileUpdate(.talk, ["level": 2])
         #expect(recorder.reports == [.volatile(.talk, ["level": 1])])
-        for _ in 0..<50 where recorder.reports.count < 2 {
-            try await Task.sleep(nanoseconds: 20_000_000)
-        }
+        try await waitUntil("trailing flush forwarded the held update") { recorder.reports.count >= 2 }
         #expect(recorder.reports == [.volatile(.talk, ["level": 1]), .volatile(.talk, ["level": 2])])
     }
 
diff --git a/Tests/OpenClawKitTests/VoiceNoteRecorderTests.swift b/Tests/OpenClawKitTests/VoiceNoteRecorderTests.swift
index 0e164ab..3cd7405 100644
--- a/Tests/OpenClawKitTests/VoiceNoteRecorderTests.swift
+++ b/Tests/OpenClawKitTests/VoiceNoteRecorderTests.swift
@@ -359,10 +359,9 @@ struct VoiceNoteRecorderTests {
         #expect(started)
         #expect(recorder.level == 0)
 
-        for _ in 0..<200 where recorder.level == 0 {
-            try await Task.sleep(nanoseconds: 3_000_000)
+        try await waitUntil("capture level published") {
+            await MainActor.run { recorder.level > 0 }
         }
-        #expect(recorder.level > 0)
 
         _ = try #require(recorder.finish())
         #expect(recorder.level == 0)
diff --git a/Tests/OpenClawLinuxRuntimeTests/AgentRuntimeExtensionsTests.swift b/Tests/OpenClawLinuxRuntimeTests/AgentRuntimeExtensionsTests.swift
index abec741..571aba6 100644
--- a/Tests/OpenClawLinuxRuntimeTests/AgentRuntimeExtensionsTests.swift
+++ b/Tests/OpenClawLinuxRuntimeTests/AgentRuntimeExtensionsTests.swift
@@ -38,7 +38,7 @@ actor SessionRoutingProvider: ModelProvider {
     }
 }
 
-@Suite("Agent runtime extensions")
+@Suite("Agent runtime extensions", .timeLimit(.minutes(1)))
 struct AgentRuntimeExtensionsTests {
     // MARK: - Exec gate
 
@@ -92,13 +92,8 @@ struct AgentRuntimeExtensionsTests {
         #expect(await broker.pending().isEmpty)
         // The fourth command escalates to a human.
         let escalated = Task { await gate.evaluate(command: "danger 4", permissionMode: .workspace, sessionKey: "w") }
-        var pending: [AgentApproval] = []
-        for _ in 0..<100 {
-            pending = await broker.pending()
-            if !pending.isEmpty { break }
-            try await Task.sleep(nanoseconds: 10_000_000)
-        }
-        let approval = try #require(pending.first)
+        try await waitUntil("escalated approval pending") { await !broker.pending().isEmpty }
+        let approval = try #require(await broker.pending().first)
         #expect(approval.presentation.warningText?.contains("Escalated") == true)
         _ = try await broker.resolve(id: approval.id, decision: .allowOnce)
         #expect(await escalated.value == .allow(source: .human))
@@ -108,9 +103,7 @@ struct AgentRuntimeExtensionsTests {
             throw ReviewDown()
         })
         let asking = Task { await failing.evaluate(command: "anything", permissionMode: .workspace, sessionKey: "f") }
-        for _ in 0..<100 where await broker.pending().isEmpty {
-            try await Task.sleep(nanoseconds: 10_000_000)
-        }
+        try await waitUntil("failed-review approval pending") { await !broker.pending().isEmpty }
         let failed = try #require(await broker.pending().first)
         #expect(failed.presentation.warningText?.contains("Automatic review failed") == true)
         await broker.cancel(runID: "none")
@@ -132,8 +125,7 @@ struct AgentRuntimeExtensionsTests {
 
     // MARK: - Sub-agents and ledger
 
-    // Bounded so a lost wake-up fails fast instead of hanging the suite.
-    @Test(.timeLimit(.minutes(1)))
+    @Test
     func spawnAnnouncesCompletionAndWakesTheYieldedParent() async throws {
         let provider = SessionRoutingProvider()
         let store = SessionStore(fileURL: FileManager.default.temporaryDirectory.appendingPathComponent("sub-\(UUID().uuidString)/sessions.json"))
@@ -148,7 +140,8 @@ struct AgentRuntimeExtensionsTests {
         let manager = SubagentManager(runtime: runtime, ledger: ledger)
         await manager.registerTools()
 
-        let result = try await runtime.run(AgentRunRequest(sessionKey: "agent:main:main", prompt: "delegate"), timeoutMs: 10_000)
+        // The run ends on cancellation, so the hour-long timeout leaves the time limit as the only bound.
+        let result = try await runtime.run(AgentRunRequest(sessionKey: "agent:main:main", prompt: "delegate"), timeoutMs: 3_600_000)
         #expect(result.toolResults.map(\.name) == ["sessions_spawn", "sessions_yield"])
         let spawnDetails = try #require(result.toolResults.first?.output.details?.dictionaryValue)
         #expect(spawnDetails["status"] == AnyCodable("accepted"))
@@ -157,20 +150,15 @@ struct AgentRuntimeExtensionsTests {
         #expect(await store.recordForKey(childKey)?.spawnedBy == "agent:main:main")
         #expect(await store.recordForKey(childKey)?.spawnDepth == 1)
 
-        var prompts: [String] = []
-        for _ in 0..<300 {
-            prompts = await provider.prompts()
-            if prompts.contains(where: { $0.contains("child result") }) { break }
-            try await Task.sleep(nanoseconds: 10_000_000)
+        try await waitUntil("parent woke with the child result") {
+            await provider.prompts().contains { $0.contains("child result") }
         }
-        #expect(prompts.contains { $0.contains("[Subagent completion]") && $0.contains("child result") })
+        #expect(await provider.prompts().contains { $0.contains("[Subagent completion]") && $0.contains("child result") })
 
         let tasks = await ledger.list().tasks
         #expect(tasks.count == 1)
         #expect(tasks.first?.kind == .subagent)
-        for _ in 0..<100 where await ledger.list().tasks.first?.status.isTerminal != true {
-            try await Task.sleep(nanoseconds: 10_000_000)
-        }
+        try await waitUntil("sub-agent task terminal") { await ledger.list().tasks.first?.status.isTerminal == true }
         #expect(await ledger.list().tasks.first?.status == .completed)
         #expect(await ledger.list(statuses: [.running]).tasks.isEmpty)
         let children = await manager.children(of: "agent:main:main")
@@ -185,7 +173,7 @@ struct AgentRuntimeExtensionsTests {
         }
     }
 
-    @Test(.timeLimit(.minutes(1)))
+    @Test
     func spawnParamsRejectUnsupportedOptionsAndKillWorks() async throws {
         #expect(throws: SubagentError.self) { try SubagentSpawnParams.parse(["task": AnyCodable("x"), "runtime": AnyCodable("acp")]) }
         #expect(throws: SubagentError.self) { try SubagentSpawnParams.parse(["task": AnyCodable("x"), "visible": AnyCodable(true)]) }
@@ -212,7 +200,7 @@ struct AgentRuntimeExtensionsTests {
         #expect(await manager.resolve(target: "1", parentSessionKey: "p").count == 1)
         let killed = try await manager.kill(target: "last", parentSessionKey: "p")
         #expect(killed.map(\.status) == ["killed"])
-        #expect(await runtime.wait(runID: record.runID, timeoutMs: 2_000)?.status == "error")
+        #expect(try await awaitCancellable("killed run finished") { await runtime.wait(runID: record.runID) }?.status == "error")
     }
 
     @Test
@@ -344,7 +332,7 @@ struct AgentRuntimeExtensionsTests {
             )
         )
         let runID = try #require(refresh.payload?.dictionaryValue?["runId"]?.stringValue)
-        #expect(await runtime.wait(runID: runID, timeoutMs: 5_000)?.status == "ok")
+        #expect(try await awaitCancellable("refresh run finished") { await runtime.wait(runID: runID) }?.status == "ok")
         // The refresh instruction is hidden from chat history.
         #expect(try await runtime.history(sessionKey: "pc").map(\.role) == ["assistant"])
     }
diff --git a/Tests/OpenClawLinuxRuntimeTests/ApprovalQuestionBrokerTests.swift b/Tests/OpenClawLinuxRuntimeTests/ApprovalQuestionBrokerTests.swift
index 348b43b..89b3974 100644
--- a/Tests/OpenClawLinuxRuntimeTests/ApprovalQuestionBrokerTests.swift
+++ b/Tests/OpenClawLinuxRuntimeTests/ApprovalQuestionBrokerTests.swift
@@ -5,7 +5,7 @@ import OpenClawModels
 import OpenClawProtocol
 @testable import OpenClawAgents
 
-@Suite("Approval and question brokers")
+@Suite("Approval and question brokers", .timeLimit(.minutes(1)))
 struct ApprovalQuestionBrokerTests {
     // MARK: - Approvals
 
@@ -56,7 +56,7 @@ struct ApprovalQuestionBrokerTests {
     func expiryFailsClosed() async throws {
         let broker = ApprovalBroker()
         let pending = await broker.request(presentation: .exec(commandText: "rm -rf build"), timeoutMs: 30)
-        let result = try await broker.waitDecision(id: pending.id, timeoutMs: 2_000)
+        let result = try await awaitCancellable("approval expired") { try await broker.waitDecision(id: pending.id) }
         #expect(result.state == .expired)
         #expect(result.reason == .timeout)
         #expect(result.isAllowed == false)
@@ -76,13 +76,8 @@ struct ApprovalQuestionBrokerTests {
         let waiter = Task {
             await broker.requestAndWait(presentation: .exec(commandText: "git status"), agentID: "main", grantKey: key)
         }
-        var pendingID: String?
-        for _ in 0..<50 {
-            pendingID = await broker.pending().first?.id
-            if pendingID != nil { break }
-            try await Task.sleep(nanoseconds: 10_000_000)
-        }
-        let id = try #require(pendingID)
+        try await waitUntil("grant approval pending") { await !broker.pending().isEmpty }
+        let id = try #require(await broker.pending().first?.id)
         _ = try await broker.resolve(id: id, decision: .allowAlways, grantExpiresInDays: 7)
         #expect(await waiter.value.isAllowed)
 
@@ -135,13 +130,11 @@ struct ApprovalQuestionBrokerTests {
             hooks: AgentLoopHooks(beforeToolCall: { _ in .requireApproval(AgentToolApprovalRequest(title: "Echo", description: "Allow?")) })
         )
         let runID = await runtime.start(AgentRunRequest(runID: "wait-approval", sessionKey: "w", prompt: "go"), streaming: false)
-        for _ in 0..<100 where await runtime.approvals.pending().isEmpty {
-            try await Task.sleep(nanoseconds: 10_000_000)
-        }
+        try await waitUntil("tool approval pending") { await !runtime.approvals.pending().isEmpty }
         let pending = try #require(await runtime.approvals.pending().first)
         #expect(pending.runID == runID)
         await runtime.abort(runID: runID)
-        let result = try #require(await runtime.wait(runID: runID, timeoutMs: 2_000))
+        let result = try #require(await awaitCancellable("aborted run finished") { await runtime.wait(runID: runID) })
         #expect(result.status == "error")
         #expect(await runtime.approvals.get(id: pending.id)?.state == .cancelled)
     }
@@ -208,13 +201,8 @@ struct ApprovalQuestionBrokerTests {
         let asking = Task {
             try await tool.invoke(AgentToolInvocation(arguments: arguments, context: AgentToolInvocationContext(runID: "r", sessionKey: "chat")), update: nil)
         }
-        var questionID: String?
-        for _ in 0..<100 {
-            questionID = await broker.pendingQuestion(sessionKey: "chat")?.id
-            if questionID != nil { break }
-            try await Task.sleep(nanoseconds: 10_000_000)
-        }
-        let id = try #require(questionID)
+        try await waitUntil("question pending") { await broker.pendingQuestion(sessionKey: "chat") != nil }
+        let id = try #require(await broker.pendingQuestion(sessionKey: "chat")?.id)
 
         // One pending question per session.
         let second = try await tool.invoke(AgentToolInvocation(arguments: arguments, context: AgentToolInvocationContext(sessionKey: "chat")), update: nil)
@@ -243,7 +231,7 @@ struct ApprovalQuestionBrokerTests {
             options: [AgentQuestionOption(label: "Yes"), AgentQuestionOption(label: "No")]
         )
         let expiring = try await broker.request(questions: [prompt], timeoutMs: 30)
-        #expect(try await broker.waitAnswer(id: expiring.id, timeoutMs: 2_000) == .expired)
+        #expect(try await awaitCancellable("question expired") { try await broker.waitAnswer(id: expiring.id) } == .expired)
 
         let cancelled = try await broker.request(questions: [prompt], runID: "run-9")
         #expect(try await broker.waitAnswer(id: cancelled.id, timeoutMs: 20) == .pending)
diff --git a/Tests/OpenClawLinuxRuntimeTests/ChannelAdaptersLinuxSmokeTests.swift b/Tests/OpenClawLinuxRuntimeTests/ChannelAdaptersLinuxSmokeTests.swift
index 53831f4..46f8ba8 100644
--- a/Tests/OpenClawLinuxRuntimeTests/ChannelAdaptersLinuxSmokeTests.swift
+++ b/Tests/OpenClawLinuxRuntimeTests/ChannelAdaptersLinuxSmokeTests.swift
@@ -69,14 +69,22 @@ struct ChannelAdaptersLinuxSmokeTests {
 
     @Test
     func a2aRoundTripCompletesTaskWithReply() async throws {
-        let config = A2AChannelConfig(enabled: true, replyTimeoutMs: 5_000, peers: ["peer": A2APeerConfig(token: "secret")])
+        // The longest reply timeout (10 min) outlasts the time limit; the task-store wait ignores
+        // cancellation, so the request is awaited through awaitCancellable.
+        let config = A2AChannelConfig(
+            enabled: true,
+            replyTimeoutMs: A2AChannelConfig.replyTimeoutRangeMs.upperBound,
+            peers: ["peer": A2APeerConfig(token: "secret")]
+        )
         let adapter = A2AChannelAdapter(config: config)
         await adapter.setInboundHandler { message in
             try? await adapter.send(OutboundMessage(channel: .a2a, peerID: message.peerID, text: "done"))
         }
         try await adapter.start()
         let request = #"{"jsonrpc":"2.0","id":1,"method":"SendMessage","params":{"message":{"role":"ROLE_USER","contextId":"c1","parts":[{"text":"task"}]}}}"#
-        let response = await adapter.handleHTTP(method: "POST", path: "/a2a/v1", headers: ["Authorization": "Bearer secret"], body: Data(request.utf8))
+        let response = try await awaitCancellable("A2A task completed") {
+            await adapter.handleHTTP(method: "POST", path: "/a2a/v1", headers: ["Authorization": "Bearer secret"], body: Data(request.utf8))
+        }
         let decoded = try JSONDecoder().decode(A2AJSONRPCResponse<A2ASendMessageResult>.self, from: response.body)
         #expect(decoded.result?.task?.status.state == .completed)
         #expect(decoded.result?.task?.replyText == "done")
@@ -106,7 +114,8 @@ struct ChannelAdaptersLinuxSmokeTests {
         try FileManager.default.setAttributes([.posixPermissions: 0o755], ofItemAtPath: script.path)
         let client = IMsgRPCClient(pipe: IMsgProcessPipe(cliPath: script.path))
         try await client.start()
-        let result = try await client.request("ping", timeoutMs: 10_000)
+        // No request timeout (0); the reply wait ignores cancellation, so it goes through awaitCancellable.
+        let result = try await awaitCancellable("imsg ping answered") { try await client.request("ping", timeoutMs: 0) }
         #expect(result.dictionaryValue?["ok"]?.boolValue == true)
         await client.stop()
     }
diff --git a/Tests/OpenClawLinuxRuntimeTests/ExecApprovalHardeningTests.swift b/Tests/OpenClawLinuxRuntimeTests/ExecApprovalHardeningTests.swift
index 733bcac..726ea7a 100644
--- a/Tests/OpenClawLinuxRuntimeTests/ExecApprovalHardeningTests.swift
+++ b/Tests/OpenClawLinuxRuntimeTests/ExecApprovalHardeningTests.swift
@@ -5,15 +5,10 @@ import OpenClawProtocol
 @testable import OpenClawAgents
 
 /// Exec approval gate and grant-key hardening (allow-always scope, raw allowlist text, closure rules).
-@Suite("Exec approval hardening")
+@Suite("Exec approval hardening", .timeLimit(.minutes(1)))
 struct ExecApprovalHardeningTests {
     private func firstPending(_ broker: ApprovalBroker) async throws -> AgentApproval {
-        for _ in 0..<300 {
-            if let approval = await broker.pending().first {
-                return approval
-            }
-            try await Task.sleep(nanoseconds: 10_000_000)
-        }
+        try await waitUntil("exec approval pending") { await !broker.pending().isEmpty }
         return try #require(await broker.pending().first)
     }
 
diff --git a/Tests/OpenClawLinuxRuntimeTests/GatewayRunLifecycleTests.swift b/Tests/OpenClawLinuxRuntimeTests/GatewayRunLifecycleTests.swift
index 6828211..cff332e 100644
--- a/Tests/OpenClawLinuxRuntimeTests/GatewayRunLifecycleTests.swift
+++ b/Tests/OpenClawLinuxRuntimeTests/GatewayRunLifecycleTests.swift
@@ -8,7 +8,7 @@ import OpenClawProtocol
 
 /// Built-in run tracking, client timeouts, pagination cursors, run-id collisions and
 /// `progressCard.put` validation (2026.3.0 FX1 review fixes).
-@Suite("Gateway run lifecycle")
+@Suite("Gateway run lifecycle", .timeLimit(.minutes(1)))
 struct GatewayRunLifecycleTests {
     typealias Harness = GatewayServerTestHarness
 
@@ -33,22 +33,24 @@ struct GatewayRunLifecycleTests {
 
     @Test
     func builtinAgentWaitTimesOutWithoutWaitingForTheRun() async throws {
+        // The run only ends when aborted, so a wait that ignored its timeout would hang until the time limit.
         let server = Self.bareServer("lifecycle-wait-timeout") { request in
-            try await Task.sleep(nanoseconds: 10_000_000_000)
+            try await Task.sleep(nanoseconds: 3_600_000_000_000)
             return Self.ok(request)
         }
         let accepted = try Harness.payload(await Harness.call(server, "agent", ["message": AnyCodable("slow"), "idempotencyKey": AnyCodable("slow-1")]))
         #expect(accepted["runId"] == AnyCodable("slow-1"))
-        let started = Date()
-        let waited = try Harness.payload(await Harness.call(server, "agent.wait", ["runId": AnyCodable("slow-1"), "timeoutMs": AnyCodable(50)]))
+        let waited = try Harness.payload(try await awaitCancellable("timed-out agent.wait returned") {
+            await Harness.call(server, "agent.wait", ["runId": AnyCodable("slow-1"), "timeoutMs": AnyCodable(50)])
+        })
         #expect(waited["status"] == AnyCodable("timeout"))
-        // Well before the 10 s run ends (generous margin for loaded CI machines).
-        #expect(Date().timeIntervalSince(started) < 5)
         // The run keeps being tracked after a timed-out wait.
         #expect(await server.trackedRuns["slow-1"] != nil)
         let aborted = try Harness.payload(await Harness.call(server, "sessions.abort", ["runId": AnyCodable("slow-1")]))
         #expect(aborted["status"] == AnyCodable("aborted"))
-        let final = try Harness.payload(await Harness.call(server, "agent.wait", ["runId": AnyCodable("slow-1"), "timeoutMs": AnyCodable(5_000)]))
+        let final = try Harness.payload(try await awaitCancellable("aborted run reported") {
+            await Harness.call(server, "agent.wait", ["runId": AnyCodable("slow-1")])
+        })
         #expect(final["status"] == AnyCodable("error"))
         #expect(final["error"] == AnyCodable("aborted"))
     }
@@ -62,7 +64,9 @@ struct GatewayRunLifecycleTests {
             return Self.ok(request)
         }
         _ = await Harness.call(server, "agent", ["message": AnyCodable("explode"), "idempotencyKey": AnyCodable("boom-1")])
-        let failed = try Harness.payload(await Harness.call(server, "agent.wait", ["runId": AnyCodable("boom-1"), "timeoutMs": AnyCodable(5_000)]))
+        let failed = try Harness.payload(try await awaitCancellable("failed run reported") {
+            await Harness.call(server, "agent.wait", ["runId": AnyCodable("boom-1")])
+        })
         #expect(failed["status"] == AnyCodable("error"))
         #expect(failed["error"] == AnyCodable("exploded"))
         #expect(failed["endedAt"] != nil)
@@ -71,9 +75,7 @@ struct GatewayRunLifecycleTests {
         _ = await Harness.call(server, "sessions.send", [
             "key": AnyCodable("agent:main:quick"), "message": AnyCodable("quick"), "idempotencyKey": AnyCodable("quick-1"),
         ])
-        for _ in 0..<200 where await server.completedRuns["quick-1"] == nil {
-            try await Task.sleep(nanoseconds: 5_000_000)
-        }
+        try await waitUntil("unwatched run completed") { await server.completedRuns["quick-1"] != nil }
         #expect(await server.agentRuns.isEmpty)
         #expect(await server.trackedRuns.isEmpty)
         let abortByKey = try Harness.payload(await Harness.call(server, "sessions.abort", ["key": AnyCodable("agent:main:quick")]))
@@ -169,7 +171,9 @@ struct GatewayRunLifecycleTests {
             "key": AnyCodable("agent:main:main"), "message": AnyCodable("hi"),
         ]))
         let runID = try #require(sent["runId"]?.stringValue)
-        _ = await Harness.call(stack.server, "agent.wait", ["runId": AnyCodable(runID), "timeoutMs": AnyCodable(5_000)])
+        _ = try await awaitCancellable("sessions.send run finished") {
+            await Harness.call(stack.server, "agent.wait", ["runId": AnyCodable(runID)])
+        }
         let task = await ledger.create(kind: .tool, sessionKey: "agent:main:main")
 
         for method in ["tasks.list", "approval.history"] {
@@ -205,9 +209,11 @@ struct GatewayRunLifecycleTests {
 
     @Test
     func reusedIdempotencyKeysDoNotStartASecondRunUnderTheSameID() async throws {
+        // The first run stays in flight until the duplicate sends are refused.
+        let inFlight = AsyncGate()
         let stack = await Harness.runtimeStack("lifecycle-run-ids", turns: [
             { _ in
-                try await Task.sleep(nanoseconds: 3_000_000_000)
+                await inFlight.wait()
                 return ModelGenerationResponse(text: "first", providerID: "scripted")
             },
         ], fallback: ScriptedToolProvider.text("later"))
@@ -225,7 +231,10 @@ struct GatewayRunLifecycleTests {
         ]))
         #expect(agent["status"] == AnyCodable("in_flight"))
         #expect(await stack.runtime.activeRunIDs() == ["msg-1"])
-        let waited = try Harness.payload(await Harness.call(stack.server, "agent.wait", ["runId": AnyCodable("msg-1"), "timeoutMs": AnyCodable(5_000)]))
+        await inFlight.open()
+        let waited = try Harness.payload(try await awaitCancellable("first run finished") {
+            await Harness.call(stack.server, "agent.wait", ["runId": AnyCodable("msg-1")])
+        })
         #expect(waited["output"] == AnyCodable("first"))
     }
 
diff --git a/Tests/OpenClawLinuxRuntimeTests/MCPHardeningTests.swift b/Tests/OpenClawLinuxRuntimeTests/MCPHardeningTests.swift
index d092169..9fa9949 100644
--- a/Tests/OpenClawLinuxRuntimeTests/MCPHardeningTests.swift
+++ b/Tests/OpenClawLinuxRuntimeTests/MCPHardeningTests.swift
@@ -370,11 +370,10 @@ struct MCPHardeningTests {
         feed.yield(Data(": ping\n\n".utf8))
         let transport = MCPLegacySSETransport(url: URL(string: "https://mcp.example.com/mcp")!, http: http)
         let client = MCPClient(serverName: "legacy", transport: transport, connectionTimeoutMs: 100)
-        let started = Date()
+        // The error names the 100 ms deadline; without one the stream never ends and the time limit fails the test.
         await #expect(throws: MCPTransportError.timeout(method: "connect", milliseconds: 100)) {
             try await client.connect()
         }
-        #expect(Date().timeIntervalSince(started) < 2)
         await #expect(throws: MCPTransportError.closed("transport closed")) {
             try await transport.send(.notification(method: "ping", params: nil))
         }
diff --git a/Tests/OpenClawLinuxRuntimeTests/MCPStdioTransportTests.swift b/Tests/OpenClawLinuxRuntimeTests/MCPStdioTransportTests.swift
index 47132b5..57a86a2 100644
--- a/Tests/OpenClawLinuxRuntimeTests/MCPStdioTransportTests.swift
+++ b/Tests/OpenClawLinuxRuntimeTests/MCPStdioTransportTests.swift
@@ -89,9 +89,8 @@ struct MCPStdioTransportTests {
         )
         try await transport.start()
         let pid = try #require(await transport.processIdentifier)
-        let startedAt = Date()
         await transport.close()
-        #expect(Date().timeIntervalSince(startedAt) < 5)
+        // The server traps TERM, so it is gone only if close escalated to KILL.
         #expect(kill(pid, 0) != 0)
     }
 
@@ -111,20 +110,26 @@ struct MCPStdioTransportTests {
 
     @Test
     func closeDoesNotWaitForAWriteBlockedOnAFullPipe() async throws {
-        // The server never reads stdin, so a 200 KB message fills the pipe and the write blocks.
+        // The server never reads stdin, so a 200 KB message fills the pipe and the write blocks. The
+        // server outlives the time limit: a close that waited for the write would hang until the
+        // cancellation handler below kills the server.
         let transport = try MCPStdioTransport(
             serverName: "deaf",
-            config: self.config("exec sleep 5"),
+            config: self.config("exec sleep 3600"),
             allowlist: ExecCommandAllowlist(patterns: ["/bin/*", "/usr/bin/*"]),
             shutdownGraceSeconds: 0.2
         )
         try await transport.start()
+        let pid = try #require(await transport.processIdentifier)
         let big = MCPJSONRPCMessage.request(id: .int(1), method: "tools/call", params: AnyCodable(["blob": AnyCodable(String(repeating: "x", count: 200_000))]))
         let pending = Task { try await transport.send(big) }
         try await Task.sleep(nanoseconds: 200_000_000)
-        let startedAt = Date()
-        await transport.close()
-        #expect(Date().timeIntervalSince(startedAt) < 3, "close escalates without waiting for the blocked write")
+        await withTaskCancellationHandler {
+            await transport.close()
+        } onCancel: {
+            _ = kill(pid, SIGKILL)
+        }
+        #expect(kill(pid, 0) != 0, "close escalates without waiting for the blocked write")
         // Once the child is gone the blocked write fails with EPIPE (no SIGPIPE crash).
         await #expect(throws: MCPTransportError.self) {
             try await pending.value
diff --git a/Tests/OpenClawLinuxRuntimeTests/RuntimeIntegrationWiringTests.swift b/Tests/OpenClawLinuxRuntimeTests/RuntimeIntegrationWiringTests.swift
index 5892f24..d2f1b4f 100644
--- a/Tests/OpenClawLinuxRuntimeTests/RuntimeIntegrationWiringTests.swift
+++ b/Tests/OpenClawLinuxRuntimeTests/RuntimeIntegrationWiringTests.swift
@@ -8,7 +8,7 @@ import OpenClawProtocol
 import OpenClawSkills
 @testable import OpenClawAgents
 
-@Suite("Runtime integration wiring")
+@Suite("Runtime integration wiring", .timeLimit(.minutes(1)))
 struct RuntimeIntegrationWiringTests {
     private static func document(_ json: String) throws -> OpenClawConfigDocument {
         try OpenClawConfigDocument.decode(Data(json.utf8))
@@ -433,9 +433,7 @@ struct RuntimeIntegrationWiringTests {
         )
         let manager = SubagentManager(runtime: runtime)
         let record = try await manager.spawn(SubagentSpawnParams(task: "summarize", label: "Summary"), parentSessionKey: "agent:main:main")
-        for _ in 0..<300 where await !log.names.contains("subagent_ended") {
-            try await Task.sleep(nanoseconds: 10_000_000)
-        }
+        try await waitUntil("subagent_ended hook ran") { await log.names.contains("subagent_ended") }
         #expect(await log.names == ["subagent_spawned", "subagent_ended"])
         let spawned = try #require(await log.first(.subagentSpawned))
         #expect(spawned["childSessionKey"]?.stringValue == record.childSessionKey)
diff --git a/Tests/OpenClawLinuxRuntimeTests/SignInWithChatGPTSessionTests.swift b/Tests/OpenClawLinuxRuntimeTests/SignInWithChatGPTSessionTests.swift
index a43aaf0..5091d87 100644
--- a/Tests/OpenClawLinuxRuntimeTests/SignInWithChatGPTSessionTests.swift
+++ b/Tests/OpenClawLinuxRuntimeTests/SignInWithChatGPTSessionTests.swift
@@ -392,7 +392,8 @@ struct SignInWithChatGPTSessionTests {
         let server = SIWCFakeServer()
         let session = SIWCTest.makeSession(server: server, clock: SIWCTestClock(), callbackPort: 0)
         let browser = LoopbackCallingBrowser()
-        let result = try await session.signIn(using: browser, reauthenticating: nil, consent: .automatic, timeout: 20, secrets: SIWCTest.secrets)
+        // Sign-in ends on cancellation, so an hour-long timeout leaves the suite's time limit as the bound.
+        let result = try await session.signIn(using: browser, reauthenticating: nil, consent: .automatic, timeout: 3_600, secrets: SIWCTest.secrets)
         #expect(result.account.subject == "user-abc")
         #expect(browser.presentedURL.map { SIWCTest.queryItems($0)["client_id"] } == "dynamic_agent_client")
         #expect(browser.dismissCount >= 1)
@@ -408,7 +409,7 @@ struct SignInWithChatGPTSessionTests {
         let server = SIWCFakeServer()
         let session = SIWCTest.makeSession(server: server, clock: SIWCTestClock(), callbackPort: 0)
         let browser = LoopbackCallingBrowser(closesAfterCallback: true)
-        let result = try await session.signIn(using: browser, reauthenticating: nil, consent: .automatic, timeout: 20, secrets: SIWCTest.secrets)
+        let result = try await session.signIn(using: browser, reauthenticating: nil, consent: .automatic, timeout: 3_600, secrets: SIWCTest.secrets)
         #expect(result.account.subject == "user-abc")
     }
 
@@ -417,13 +418,13 @@ struct SignInWithChatGPTSessionTests {
         let server = SIWCFakeServer()
         let session = SIWCTest.makeSession(server: server, clock: SIWCTestClock(), callbackPort: 0)
         await #expect(throws: SignInWithChatGPTError.cancelled) {
-            try await session.signIn(using: CancellingBrowser(), timeout: 20)
+            try await session.signIn(using: CancellingBrowser(), timeout: 3_600)
         }
         await #expect(throws: SignInWithChatGPTError.timedOut) {
             try await session.signIn(using: SignInWithChatGPTExternalBrowser { _ in true }, timeout: 1)
         }
         await #expect(throws: OpenClawCoreError.self) {
-            try await session.signIn(using: SignInWithChatGPTExternalBrowser { _ in false }, timeout: 20)
+            try await session.signIn(using: SignInWithChatGPTExternalBrowser { _ in false }, timeout: 3_600)
         }
         #expect(try await session.accounts().isEmpty)
     }
diff --git a/Tests/OpenClawLinuxRuntimeTests/SpotlightMemoryTests.swift b/Tests/OpenClawLinuxRuntimeTests/SpotlightMemoryTests.swift
index 64f8c86..b247c9e 100644
--- a/Tests/OpenClawLinuxRuntimeTests/SpotlightMemoryTests.swift
+++ b/Tests/OpenClawLinuxRuntimeTests/SpotlightMemoryTests.swift
@@ -10,12 +10,31 @@ import OpenClawProtocol
 final class FakeSpotlightStore: SpotlightItemIndexing, @unchecked Sendable {
     private let lock = NSLock()
     private var items: [String: String] = [:]
-    private let hangs: Bool
+    private var hangs: Bool
+    private var stalled: [@Sendable ((any Error)?) -> Void] = []
 
     init(hangs: Bool = false) {
         self.hangs = hangs
     }
 
+    /// Answers the calls a hanging store left pending, and every later call.
+    func stopHanging() {
+        let stalled = self.lock.withLock {
+            self.hangs = false
+            defer { self.stalled.removeAll() }
+            return self.stalled
+        }
+        for completion in stalled { completion(nil) }
+    }
+
+    /// Keeps `completion` pending when the store hangs.
+    private func stalls(_ completion: @escaping @Sendable ((any Error)?) -> Void) -> Bool {
+        self.lock.withLock {
+            if self.hangs { self.stalled.append(completion) }
+            return self.hangs
+        }
+    }
+
     var isAvailable: Bool {
         true
     }
@@ -28,7 +47,7 @@ final class FakeSpotlightStore: SpotlightItemIndexing, @unchecked Sendable {
     }
 
     func indexItems(_ items: [CSSearchableItem], completion: @escaping @Sendable ((any Error)?) -> Void) {
-        guard !self.hangs else { return }
+        guard !self.stalls(completion) else { return }
         self.lock.lock()
         for item in items { self.items[item.uniqueIdentifier] = item.domainIdentifier ?? "" }
         self.lock.unlock()
@@ -36,7 +55,7 @@ final class FakeSpotlightStore: SpotlightItemIndexing, @unchecked Sendable {
     }
 
     func deleteItems(identifiers: [String], completion: @escaping @Sendable ((any Error)?) -> Void) {
-        guard !self.hangs else { return }
+        guard !self.stalls(completion) else { return }
         self.lock.lock()
         for id in identifiers { self.items.removeValue(forKey: id) }
         self.lock.unlock()
@@ -44,7 +63,7 @@ final class FakeSpotlightStore: SpotlightItemIndexing, @unchecked Sendable {
     }
 
     func deleteItems(domains: [String], completion: @escaping @Sendable ((any Error)?) -> Void) {
-        guard !self.hangs else { return }
+        guard !self.stalls(completion) else { return }
         self.lock.lock()
         self.items = self.items.filter { _, domain in
             !domains.contains { domain == $0 || domain.hasPrefix($0 + ".") }
@@ -54,6 +73,20 @@ final class FakeSpotlightStore: SpotlightItemIndexing, @unchecked Sendable {
     }
 }
 
+/// Tells a stalled test operation to stop.
+final class StallFlag: @unchecked Sendable {
+    private let lock = NSLock()
+    private var ended = false
+
+    var isEnded: Bool {
+        self.lock.withLock { self.ended }
+    }
+
+    func end() {
+        self.lock.withLock { self.ended = true }
+    }
+}
+
 @Suite("Spotlight memory backends")
 struct SpotlightMemoryTests {
     static var live: Bool {
@@ -83,17 +116,22 @@ struct SpotlightMemoryTests {
         #expect(SpotlightTimeoutRace.deadlineNanoseconds(-5) == 50_000_000)
     }
 
-    @Test
+    @Test(.timeLimit(.minutes(1)))
     func indexWritesAreBoundedWhenSpotlightNeverAnswers() async throws {
-        let index = Self.index(store: FakeSpotlightStore(hangs: true), writeTimeoutSeconds: 0.2)
-        let started = Date()
-        await #expect(throws: SpotlightTimeoutError.self) {
-            try await index.upsert([MemoryDocument(id: "a", source: .systemNote, text: "stalled write")], sessionKey: nil)
-        }
-        await #expect(throws: SpotlightTimeoutError.self) {
-            try await index.deleteAll()
+        let store = FakeSpotlightStore(hangs: true)
+        let index = Self.index(store: store, writeTimeoutSeconds: 0.2)
+        // The errors name the 0.2 s deadline. A write without one would hang until the time limit,
+        // whose cancellation answers the stalled calls so the test can end.
+        await withTaskCancellationHandler {
+            await #expect(throws: SpotlightTimeoutError(operation: "indexSearchableItems", seconds: 0.2)) {
+                try await index.upsert([MemoryDocument(id: "a", source: .systemNote, text: "stalled write")], sessionKey: nil)
+            }
+            await #expect(throws: SpotlightTimeoutError(operation: "deleteSearchableItems(withDomainIdentifiers:)", seconds: 0.2)) {
+                try await index.deleteAll()
+            }
+        } onCancel: {
+            store.stopHanging()
         }
-        #expect(Date().timeIntervalSince(started) < 3)
     }
 
     @Test
@@ -229,18 +267,26 @@ struct SpotlightMemoryTests {
         feed.finish()
     }
 
-    @Test
-    func timeoutRaceReturnsWithoutJoiningAStalledOperation() async {
-        let started = Date()
+    @Test(.timeLimit(.minutes(1)))
+    func timeoutRaceReturnsWithoutJoiningAStalledOperation() async throws {
         // The operation ignores cancellation and never finishes on its own, like a stalled CSUserQuery.
-        let value: Int? = await SpotlightTimeoutRace.first(timeoutSeconds: 0.2) {
-            while true {
-                try? await Task.sleep(nanoseconds: 50_000_000)
+        // It stops once the test ends, or when the time limit cancels a race that waits for it.
+        let stall = StallFlag()
+        defer { stall.end() }
+        let value: Int? = await withTaskCancellationHandler {
+            await SpotlightTimeoutRace.first(timeoutSeconds: 0.2) {
+                while !stall.isEnded {
+                    try? await Task.sleep(nanoseconds: 50_000_000)
+                }
+                return 0
             }
+        } onCancel: {
+            stall.end()
         }
         #expect(value == nil)
-        #expect(Date().timeIntervalSince(started) < 2)
-        let fast: Int? = await SpotlightTimeoutRace.first(timeoutSeconds: 5) { 42 }
+        let fast: Int? = try await awaitCancellable("fast operation won the race") {
+            await SpotlightTimeoutRace.first(timeoutSeconds: 3_600) { 42 }
+        }
         #expect(fast == 42)
     }
 
diff --git a/Tests/OpenClawLinuxRuntimeTests/SubagentRuntimeHardeningTests.swift b/Tests/OpenClawLinuxRuntimeTests/SubagentRuntimeHardeningTests.swift
index 1a855f1..189fded 100644
--- a/Tests/OpenClawLinuxRuntimeTests/SubagentRuntimeHardeningTests.swift
+++ b/Tests/OpenClawLinuxRuntimeTests/SubagentRuntimeHardeningTests.swift
@@ -108,7 +108,7 @@ struct SubagentRuntimeHardeningTests {
         )
         let manager = SubagentManager(runtime: runtime)
         let child = try await manager.spawn(SubagentSpawnParams(task: "run rm -rf build and write X"), parentSessionKey: parent)
-        #expect(await runtime.wait(runID: child.runID, timeoutMs: 10_000)?.status == "ok")
+        #expect(try await awaitCancellable("child run finished") { await runtime.wait(runID: child.runID) }?.status == "ok")
 
         let childRecord = try #require(await store.recordForKey(child.childSessionKey))
         #expect(childRecord.permissionMode == .readOnly)
@@ -143,11 +143,11 @@ struct SubagentRuntimeHardeningTests {
         await manager.registerTools()
         let result = try await runtime.run(
             AgentRunRequest(sessionKey: "agent:main:main", prompt: "delegate", toolPolicy: ToolPolicy(deny: ["exec"])),
-            timeoutMs: 10_000
+            timeoutMs: 3_600_000
         )
         #expect(result.output == "parent done")
         let child = try #require(await manager.children(of: "agent:main:main").first)
-        #expect(await runtime.wait(runID: child.runID, timeoutMs: 10_000)?.status == "ok")
+        #expect(try await awaitCancellable("child run finished") { await runtime.wait(runID: child.runID) }?.status == "ok")
         #expect(await log.count("exec") == 0)
         let results = toolResults(try await runtime.history(sessionKey: child.childSessionKey))
         #expect(results.first?.text == "Tool exec is not allowed by the current tool policy")
@@ -190,20 +190,28 @@ struct SubagentRuntimeHardeningTests {
         await runtime.start(AgentRunRequest(runID: "dup", sessionKey: "two", prompt: "b"), streaming: false)
         #expect(await runtime.activeRunIDs() == ["dup"])
         await provider.release()
-        #expect(await runtime.wait(runID: "dup", timeoutMs: 10_000)?.output == "first")
+        #expect(try await awaitCancellable("joined run finished") { await runtime.wait(runID: "dup") }?.output == "first")
         #expect(await provider.count() == 1)
     }
 
     @Test(.timeLimit(.minutes(1)))
     func aStaleTimerNeverCancelsANewRunWithTheSameID() async throws {
-        let provider = GatedTextProvider(replies: ["old", "new"])
-        await provider.release()
-        let runtime = EmbeddedAgentRuntime(
-            toolRegistry: AgentToolRegistry(),
-            modelRouter: ModelRouter(defaultProviderID: provider.id, providers: [provider])
-        )
-        await runtime.start(AgentRunRequest(runID: "reuse", sessionKey: "s", prompt: "a"), timeoutMs: 200, streaming: false)
-        #expect(await runtime.wait(runID: "reuse", timeoutMs: 10_000)?.output == "old")
+        // The first run has to finish before its 200 ms timer so the timer is stale during the second
+        // run. A stalled pool can let the timer win, so start over until the first run finishes first.
+        var provider: GatedTextProvider
+        var runtime: EmbeddedAgentRuntime
+        var first: AgentRunWaitResult?
+        repeat {
+            provider = GatedTextProvider(replies: ["old", "new"])
+            await provider.release()
+            runtime = EmbeddedAgentRuntime(
+                toolRegistry: AgentToolRegistry(),
+                modelRouter: ModelRouter(defaultProviderID: provider.id, providers: [provider])
+            )
+            await runtime.start(AgentRunRequest(runID: "reuse", sessionKey: "s", prompt: "a"), timeoutMs: 200, streaming: false)
+            first = try await awaitCancellable("first run finished") { [runtime] in await runtime.wait(runID: "reuse") }
+        } while first?.status == "timeout"
+        #expect(first?.output == "old")
 
         await provider.block()
         await runtime.start(AgentRunRequest(runID: "reuse", sessionKey: "s", prompt: "b"), streaming: false)
@@ -213,7 +221,8 @@ struct SubagentRuntimeHardeningTests {
         #expect(await runtime.activeRunIDs() == ["reuse"])
         await provider.release()
         // The waiter sees the new run, not the previous run's retained result.
-        #expect(await runtime.wait(runID: "reuse", timeoutMs: 10_000)?.output == "new")
+        let second = try await awaitCancellable("second run finished") { [runtime] in await runtime.wait(runID: "reuse") }
+        #expect(second?.output == "new")
         _ = await runtime.approvals.cancel(id: approval.id)
     }
 
diff --git a/Tests/OpenClawLinuxRuntimeTests/TestAsyncHelpers.swift b/Tests/OpenClawLinuxRuntimeTests/TestAsyncHelpers.swift
index ac54535..27cb5d2 100644
--- a/Tests/OpenClawLinuxRuntimeTests/TestAsyncHelpers.swift
+++ b/Tests/OpenClawLinuxRuntimeTests/TestAsyncHelpers.swift
@@ -31,3 +31,27 @@ func recordedWaitTimeout(_ label: String) -> AsyncWaitTimeoutError {
     Issue.record(timeout)
     return timeout
 }
+
+/// Awaits `operation` in its own task, so a time-limit cancellation ends the wait even when the
+/// operation ignores cancellation.
+///
+/// Product waits such as `EmbeddedAgentRuntime.wait(runID:)`, the gateway's `agent.wait` and the
+/// approval and question brokers park on a continuation that only their own timer or the awaited
+/// event resumes. Wrapping them here lets a test wait without a timeout: a regression shows up as a
+/// hang, and the time limit turns it into an ``AsyncWaitTimeoutError`` naming `label`.
+func awaitCancellable<T: Sendable>(_ label: String, _ operation: @escaping @Sendable () async throws -> T) async throws -> T {
+    let (results, continuation) = AsyncStream<Result<T, any Error>>.makeStream(bufferingPolicy: .bufferingNewest(1))
+    let task = Task {
+        do {
+            continuation.yield(.success(try await operation()))
+        } catch {
+            continuation.yield(.failure(error))
+        }
+        continuation.finish()
+    }
+    defer { task.cancel() }
+    for await result in results {
+        return try result.get()
+    }
+    throw recordedWaitTimeout(label)
+}
diff --git a/Tests/OpenClawLinuxRuntimeTests/ToolsGatewayMethodsTests.swift b/Tests/OpenClawLinuxRuntimeTests/ToolsGatewayMethodsTests.swift
index ab2c7c8..057bd5a 100644
--- a/Tests/OpenClawLinuxRuntimeTests/ToolsGatewayMethodsTests.swift
+++ b/Tests/OpenClawLinuxRuntimeTests/ToolsGatewayMethodsTests.swift
@@ -7,7 +7,7 @@ import OpenClawProtocol
 @testable import OpenClawAgents
 
 /// `tools.catalog`, `tools.effective` and `tools.invoke` over the runtime tool registry and policy.
-@Suite("Tools gateway methods")
+@Suite("Tools gateway methods", .timeLimit(.minutes(1)))
 struct ToolsGatewayMethodsTests {
     private typealias Harness = GatewayServerTestHarness
 
@@ -164,13 +164,8 @@ struct ToolsGatewayMethodsTests {
                 "name": AnyCodable("lookup"), "args": AnyCodable(["text": AnyCodable("w")]), "confirm": AnyCodable(true),
             ]))
         }
-        var pendingID: String?
-        for _ in 0..<200 {
-            pendingID = await stack.runtime.approvals.pending().first?.id
-            if pendingID != nil { break }
-            try await Task.sleep(nanoseconds: 10_000_000)
-        }
-        let waitingID = try #require(pendingID)
+        try await waitUntil("confirmation approval pending") { await !stack.runtime.approvals.pending().isEmpty }
+        let waitingID = try #require(await stack.runtime.approvals.pending().first?.id)
         _ = await Harness.call(stack.server, "approval.resolve", ["id": AnyCodable(waitingID), "decision": AnyCodable("deny")])
         let denied = try await waiting.value
         #expect(denied["ok"] == AnyCodable(false))