fix(e2e): configure supported rolling agents with warm policy

The warm environment was overridden by each profile’s explicit per-turn agent policy. Use the production managed warm capability to select supported profiles, set their agent policy to warm, and retain per-turn coverage for every qualified native profile.

The effective-policy regression fails on the old catalog; 39 affected catalog and controller checks pass. Shared-package type declarations need a fresh build for the broad local typecheck.

Co-Authored-By: Paperclip <noreply@paperclip.ing>

tests/runner-e2e/README.md

@@ -191,6 +191,9 @@ Postgres. Each controller has an independent compiled build and cache. The copie
 app retains the checkout's `cli/` and `packages/` dependency links, including the
 development Runner vendor entry. Shared workspace dependencies mean this is
 not released-container or Railway evidence.
+The warm lane selects providers through the production managed warm-session
+capability and explicitly sets the agent policy to warm. The per-turn lane
+retains every registered provider's per-turn policy.
 The read-only recovery observer captures the live predecessor's identity just
 before handoff, then requires a new controller generation, authenticated Runner
 attachment, and new canonical events for the original execution. No fixture

tests/runner-e2e/catalog.test.ts

@@ -33,6 +33,24 @@ import {
 } from "./selectors.js";
 
 describe("runner E2E catalog", () => {
+  it("uses effective warm agent policy only for supported rolling warm providers", () => {
+    const warm = runnerMatrix.filter(cell => cell.suite.id === "rolling-web-deployment-warm");
+    const perTurn = runnerMatrix.filter(cell => cell.suite.id === "rolling-web-deployment");
+    expect(warm.some(cell => cell.profile.id === "runner-codex")).toBe(true);
+    expect(warm.some(cell => ["runner-opencode", "runner-acpx-claude", "runner-acpx-grok"].includes(cell.profile.id))).toBe(false);
+    expect(perTurn.length).toBeGreaterThan(warm.length);
+    for (const cell of [...warm, ...perTurn]) {
+      const secretRefs = Object.fromEntries([cell.profile.credential, "DAYTONA_API_KEY"].map(key => [key,
+        { type: "secret_ref" as const, secretId: "22222222-2222-4222-8222-222222222222", version: "latest" as const }]));
+      const input = { executionId: cell.id, secretRefs, daytonaImage: `fixture@sha256:${"a".repeat(64)}` };
+      const agent = cell.profile.buildAgent({ ...input, environmentId: "env-1", environmentFixtureId: "daytona", workspacePath: "/workspace" });
+      const environment = cell.environment.buildEnvironment(input);
+      const expected = cell.suite.id === "rolling-web-deployment-warm" ? "warm" : "per_turn";
+      expect(agent.adapterConfig).toMatchObject({ lifecycleMode: expected });
+      expect(environment.config).toMatchObject({ runnerLifecycleMode: expected, reuseLease: expected === "warm" });
+      expect(cell.suite.expectedMatrixSize).toBe(cell.suite.profiles.length * cell.suite.tasks.length);
+    }
+  });
   it("keeps verified artifact transfer a single explicit Daytona run", () => {
     const id = "api-artifact-transfer.runner-codex.daytona.verified-artifact-copy";
     const cells = selectRunnerExecutions(parseRunnerSelectors(["--id", id]));
@@ -312,11 +330,11 @@ describe("runner E2E catalog", () => {
     expect(localIntegrityTasks).toHaveLength(2);
     expect(openRouterBreadthTasks).toHaveLength(3);
     expect(runnerSuites.map((suite) => suite.expectedMatrixSize)).toEqual([
-      1, 9, 9, 9, 9, 4, 2, 1, 3, 2, 2, 4, 4, 4, 4, 4, 4, 8, 10, 6, 63, 12, 6, 4, 2, 4, 2, 30, 3, 16, 16, 20, 2, 1, 6, 8, 46, 2, 23, 15, 52, 6, 6, 20, 26, 52, 5, 28, 18, 2, 6, 6, 13, 10, 4, 48, 16, 10, 2, 1, 1, 4, 116, 2, 1, 2,
+      1, 9, 2, 9, 9, 4, 2, 1, 3, 2, 2, 4, 4, 4, 4, 4, 4, 8, 10, 6, 63, 12, 6, 4, 2, 4, 2, 30, 3, 16, 16, 20, 2, 1, 6, 8, 46, 2, 23, 15, 52, 6, 6, 20, 26, 52, 5, 28, 18, 2, 6, 6, 13, 10, 4, 48, 16, 10, 2, 1, 1, 4, 116, 2, 1, 2,
 
     ]);
-    expect(validateRunnerCatalog()).toHaveLength(837);
-    expect(new Set(runnerMatrix.map((entry) => entry.id)).size).toBe(837);
+    expect(validateRunnerCatalog()).toHaveLength(830);
+    expect(new Set(runnerMatrix.map((entry) => entry.id)).size).toBe(830);
     expect(
       runnerMatrix.filter((entry) => entry.suite.id === "core-compatibility"),
     ).toHaveLength(48);

tests/runner-e2e/catalog.ts

@@ -1,4 +1,5 @@
 import { ROLLING_WEB_DEPLOYMENT_SUITE, ROLLING_WEB_DEPLOYMENT_WARM_SUITE } from "./controller-rolling-deployment.js";
+import { supportsManagedNativeWarmSession } from "../../server/src/services/native-runtime/native-warm-capability.js";
 import { FEEDBACK_PERSISTENCE_SUITE, feedbackPersistenceTask, feedbackPersistenceDigest } from "./feedback-persistence.js";
 import { followupStopTask, followupStopDigest } from "./followup-stop-resume.js";
 import { finalizationRecoveryTask, finalizationRecoveryDigest } from "./finalization-recovery.js";
@@ -1252,15 +1253,24 @@ const rollingDeploymentProfiles = [...runnerProfiles, ...extendedHarnessProfiles
       return { ...agent, adapterConfig };
     } };
   });
+const rollingWarmDeploymentProfiles = rollingDeploymentProfiles.filter(profile => {
+  const backend = profile.provider === "codex" ? "codex_app_server" : profile.provider === "opencode" ? "opencode_server" : "acpx_runtime";
+  return supportsManagedNativeWarmSession({ backend, acpxAgent: "acpxAgent" in profile ? profile.acpxAgent : undefined },
+    profile.credential === "COPILOT_GITHUB_TOKEN" ? { provider: "github", method: "api_key" } : undefined);
+}).map(profile => ({ ...profile, buildAgent(input: AgentFixtureBuildInput) {
+  const agent = profile.buildAgent(input);
+  return { ...agent, adapterConfig: { ...(agent.adapterConfig as Record<string, unknown>), lifecycleMode: "warm" } };
+} }));
+
 const rollingDeploymentSuites: RunnerSuiteFixture[] = [
   { id: ROLLING_WEB_DEPLOYMENT_SUITE, environment: runnerEnvironments.find(environment => environment.id === "daytona")! },
   { id: ROLLING_WEB_DEPLOYMENT_WARM_SUITE, environment: daytonaControllerOutageEnvironment },
 ].map(({ id, environment }) => ({
   ...remoteWorkspaceStorageSuite, id, label: `Rolling control-plane web deployment (${environment.configurationKey ? "warm" : "per-turn"})`,
   description: "Overlap independent ephemeral web servers while a remote execution stays active; keep ordinary reads, saves, files and a later native continuation working.",
-  profiles: rollingDeploymentProfiles, environments: [environment],
-  expectedMatrixSize: rollingDeploymentProfiles.length * remoteWorkspaceStorageSuite.tasks.length,
-  definitionMetadata: { ...remoteWorkspaceStorageSuite.definitionMetadata, version: 4, controllerDowntimeMs: 0,
+  profiles: id === ROLLING_WEB_DEPLOYMENT_WARM_SUITE ? rollingWarmDeploymentProfiles : rollingDeploymentProfiles, environments: [environment],
+  expectedMatrixSize: (id === ROLLING_WEB_DEPLOYMENT_WARM_SUITE ? rollingWarmDeploymentProfiles : rollingDeploymentProfiles).length * remoteWorkspaceStorageSuite.tasks.length,
+  definitionMetadata: { ...remoteWorkspaceStorageSuite.definitionMetadata, version: 5, controllerDowntimeMs: 0,
     activeWorkBoundary: "foreground-180-second-cli-reliability-soak-and-live-owner-recheck",
     workspaceFileSource: environment.configurationKey ? "live" : "saved-exact-snapshot-after-per-turn-retirement",
     lifecycleCoverage: environment.configurationKey ? "warm-reusable" : "per-turn-disposable", providerCoverage: "registered-native-daytona-profiles",

tests/runner-e2e/controller-rolling-deployment.test.ts

@@ -57,7 +57,7 @@ describe("rolling controller qualification", () => {
     const [outage] = selectRunnerExecutions(parseRunnerSelectors(["--suite", "remote-workspace-storage"]));
     expect(controllerOutageForExecution(outage!, {})).toEqual({ signal: "SIGKILL", holdMs: 300_000 });
   });
-  it("requires both lifecycle modes for every production-qualified native sandbox profile", () => {
+  it("covers every qualified native profile per turn and each supported warm profile", () => {
     const executions = ["rolling-web-deployment", "rolling-web-deployment-warm"]
       .flatMap(suite => selectRunnerExecutions(parseRunnerSelectors(["--suite", suite])));
     const native = [...runnerProfiles, ...extendedHarnessProfiles, kimiRunnerProfile]
@@ -74,12 +74,14 @@ describe("rolling controller qualification", () => {
         && (!profile.qualificationCandidate || QUALIFIED_ACPX_PROFILES[profile.qualificationCandidate].qualificationStatus !== "pending"))
       .map(kind));
     expect(new Set(native.map(kind))).toEqual(registeredKinds);
+    const warmIds = new Set(runnerSuites.find(suite => suite.id === "rolling-web-deployment-warm")!.profiles.map(profile => profile.id));
     for (const profile of native) {
       const cells = executions.filter(execution => execution.profile.id === profile.id);
-      expect(cells).toHaveLength(2);
+      expect(cells.filter(cell => cell.suite.id === "rolling-web-deployment")).toHaveLength(1);
+      expect(cells).toHaveLength(warmIds.has(profile.id) ? 2 : 1);
       expect(cells.every(cell => !cell.profile.qualificationCandidate)).toBe(true);
       expect(new Set(cells.map(cell => cell.environment.configurationKey ?? "per_turn")))
-        .toEqual(new Set(["per_turn", "warm-controller-outage-v1"]));
+        .toEqual(new Set(warmIds.has(profile.id) ? ["per_turn", "warm-controller-outage-v1"] : ["per_turn"]));
     }
     expect(executions.every(execution => execution.suite.definitionMetadata?.retainedControllerContinuation === true)).toBe(true);
   });