test: give the jobs manager a shutdown budget that survives a loaded machine

internal/app/app_jobs_e2e_test.go

@@ -148,7 +148,7 @@ func TestE2E_SpawnAgentToolViaSlashCommand(t *testing.T) {
 	if err != nil {
 		t.Fatalf("jobs.NewManager: %v", err)
 	}
-	defer func() { _ = mgr.Shutdown(2 * time.Second) }()
+	defer func() { _ = mgr.Shutdown(testwait.Timeout(2 * time.Second)) }()
 
 	// Build a minimal App by hand so we don't have to stand up every
 	// Deps field. The fields we need for this test are: deps.Jobs,

internal/app/app_lifecycle_parity_test.go

@@ -7,6 +7,7 @@ import (
 	tea "github.com/charmbracelet/bubbletea"
 
 	"github.com/packetcode/packetcode/internal/agent"
+	"github.com/packetcode/packetcode/internal/testwait"
 	"github.com/packetcode/packetcode/internal/ui/components/agentview"
 )
 
@@ -28,7 +29,7 @@ func TestFirstVisibleProgressStopsThinkingSpinner(t *testing.T) {
 func TestLeftArrowOpensAgentsOnlyFromEmptyIdleInput(t *testing.T) {
 	r := newTestApp(t)
 	mgr := wireJobsManagerForSlashTest(t, r)
-	t.Cleanup(func() { _ = mgr.Shutdown(2 * time.Second) })
+	t.Cleanup(func() { _ = mgr.Shutdown(testwait.Timeout(2 * time.Second)) })
 	r.app.input.Reset()
 	r.app.handleKey(tea.KeyMsg{Type: tea.KeyLeft})
 	if !r.app.agentView.Visible() {
@@ -46,7 +47,7 @@ func TestLeftArrowOpensAgentsOnlyFromEmptyIdleInput(t *testing.T) {
 func TestAgentWorkspaceTaskPromptCanClearReturnAndSpawn(t *testing.T) {
 	r := newTestApp(t)
 	mgr := wireJobsManagerForSlashTest(t, r)
-	t.Cleanup(func() { _ = mgr.Shutdown(2 * time.Second) })
+	t.Cleanup(func() { _ = mgr.Shutdown(testwait.Timeout(2 * time.Second)) })
 
 	r.app.showAgentView()
 	r.app.handleKey(tea.KeyMsg{Type: tea.KeyRunes, Runes: []rune("n")})
@@ -84,7 +85,7 @@ func TestAgentWorkspaceTaskPromptCanClearReturnAndSpawn(t *testing.T) {
 func TestAgentWorkspaceListActionsAreNotSwallowedByTaskInput(t *testing.T) {
 	r := newTestApp(t)
 	mgr := wireJobsManagerForSlashTest(t, r)
-	t.Cleanup(func() { _ = mgr.Shutdown(2 * time.Second) })
+	t.Cleanup(func() { _ = mgr.Shutdown(testwait.Timeout(2 * time.Second)) })
 
 	_, _ = r.app.handleSpawnCommand([]string{"inspect the renderer"})
 	r.app.showAgentView()

internal/app/app_slashcmd_test.go

@@ -22,6 +22,7 @@ import (
 	"github.com/packetcode/packetcode/internal/provider"
 	"github.com/packetcode/packetcode/internal/session"
 	"github.com/packetcode/packetcode/internal/statusline"
+	"github.com/packetcode/packetcode/internal/testwait"
 	"github.com/packetcode/packetcode/internal/tools"
 	"github.com/packetcode/packetcode/internal/ui/components/agentview"
 	"github.com/packetcode/packetcode/internal/ui/components/approval"
@@ -1459,7 +1460,7 @@ func TestApp_Queue_ListDropAndClear(t *testing.T) {
 func TestApp_Agents_ListUsesBackgroundJobs(t *testing.T) {
 	r := newTestApp(t)
 	mgr := wireJobsManagerForSlashTest(t, r)
-	defer func() { _ = mgr.Shutdown(2 * time.Second) }()
+	defer func() { _ = mgr.Shutdown(testwait.Timeout(2 * time.Second)) }()
 
 	snap, spawnErr := mgr.Spawn(jobs.SpawnRequest{
 		Prompt:   "audit fixtures",
@@ -1483,7 +1484,7 @@ func TestApp_Agents_ListUsesBackgroundJobs(t *testing.T) {
 func TestApp_Agents_DetailOpensJobsPanel(t *testing.T) {
 	r := newTestApp(t)
 	mgr := wireJobsManagerForSlashTest(t, r)
-	defer func() { _ = mgr.Shutdown(2 * time.Second) }()
+	defer func() { _ = mgr.Shutdown(testwait.Timeout(2 * time.Second)) }()
 
 	snap, spawnErr := mgr.Spawn(jobs.SpawnRequest{
 		Prompt:   "inspect flaky test",
@@ -1503,7 +1504,7 @@ func TestApp_Agents_DetailOpensJobsPanel(t *testing.T) {
 func TestApp_Agents_NotFoundUsesAgentLabel(t *testing.T) {
 	r := newTestApp(t)
 	mgr := wireJobsManagerForSlashTest(t, r)
-	defer func() { _ = mgr.Shutdown(2 * time.Second) }()
+	defer func() { _ = mgr.Shutdown(testwait.Timeout(2 * time.Second)) }()
 
 	r.app.handleSlashCommand("agents", []string{"missing"}, "/agents missing")
 	convContains(t, r.app, "[agent:missing not found]")
@@ -1512,7 +1513,7 @@ func TestApp_Agents_NotFoundUsesAgentLabel(t *testing.T) {
 func TestApp_Agents_ViewDoesNotOverrideJobsPanelOverlay(t *testing.T) {
 	r := newTestApp(t)
 	mgr := wireJobsManagerForSlashTest(t, r)
-	defer func() { _ = mgr.Shutdown(2 * time.Second) }()
+	defer func() { _ = mgr.Shutdown(testwait.Timeout(2 * time.Second)) }()
 
 	snap, spawnErr := mgr.Spawn(jobs.SpawnRequest{
 		Prompt:   "inspect overlay",
@@ -1646,7 +1647,7 @@ func TestApp_ResizeRecomposesBuiltInStatusLine(t *testing.T) {
 func TestApp_HandleJobUpdate_IgnoresStaleSeqAndDedupesTerminal(t *testing.T) {
 	r := newTestApp(t)
 	mgr := wireJobsManagerForSlashTest(t, r)
-	defer func() { _ = mgr.Shutdown(2 * time.Second) }()
+	defer func() { _ = mgr.Shutdown(testwait.Timeout(2 * time.Second)) }()
 
 	snap, spawnErr := mgr.Spawn(jobs.SpawnRequest{
 		Prompt:   "finish once",

internal/jobs/abandoned_test.go

@@ -9,6 +9,7 @@ import (
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 
+	"github.com/packetcode/packetcode/internal/testwait"
 	"github.com/packetcode/packetcode/internal/tools"
 )
 
@@ -133,7 +134,7 @@ func TestCancel_RecordsRequestAndStaysCancelled(t *testing.T) {
 	// loadPersistedJobs reconciles anything still marked running — so reading
 	// the directory under a live manager would race the flush and rewrite the
 	// very record under test.
-	require.NoError(t, mgr.Shutdown(2*time.Second))
+	require.NoError(t, mgr.Shutdown(testwait.Timeout(2*time.Second)))
 
 	reloaded, _, unread, lerr := loadPersistedJobs(jobsDir, "")
 	require.NoError(t, lerr)

internal/jobs/manager_test.go

@@ -18,6 +18,7 @@ import (
 	"github.com/packetcode/packetcode/internal/config"
 	"github.com/packetcode/packetcode/internal/permissions"
 	"github.com/packetcode/packetcode/internal/provider"
+	"github.com/packetcode/packetcode/internal/testwait"
 	"github.com/packetcode/packetcode/internal/tools"
 )
 
@@ -322,7 +323,7 @@ func TestManager_Shutdown_Persists(t *testing.T) {
 		return running == N
 	})
 
-	require.NoError(t, mgr.Shutdown(2*time.Second))
+	require.NoError(t, mgr.Shutdown(testwait.Timeout(2*time.Second)))
 
 	for _, id := range ids {
 		snap, ok := mgr.Get(id)

internal/jobs/persistence_test.go

@@ -10,6 +10,8 @@ import (
 
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
+
+	"github.com/packetcode/packetcode/internal/testwait"
 )
 
 // TestSaveSnapshot_AtomicWrite confirms the temp-file-then-rename
@@ -175,7 +177,7 @@ func TestManagerPersistenceDebouncesNonterminalUpdates(t *testing.T) {
 	dir := t.TempDir()
 	mgr, _, err := NewManager(Config{JobsDir: dir})
 	require.NoError(t, err)
-	t.Cleanup(func() { _ = mgr.Shutdown(time.Second) })
+	t.Cleanup(func() { _ = mgr.Shutdown(testwait.Timeout(time.Second)) })
 	mgr.persistDelay = time.Hour
 
 	for seq := int64(1); seq <= 3; seq++ {
@@ -197,7 +199,7 @@ func TestManagerPersistenceTerminalFlushIsSynchronous(t *testing.T) {
 	dir := t.TempDir()
 	mgr, _, err := NewManager(Config{JobsDir: dir})
 	require.NoError(t, err)
-	t.Cleanup(func() { _ = mgr.Shutdown(time.Second) })
+	t.Cleanup(func() { _ = mgr.Shutdown(testwait.Timeout(time.Second)) })
 	mgr.persistDelay = time.Hour
 	require.NoError(t, mgr.savePersistedSnapshot(persistedJob{ID: "terminal", State: "running", Seq: 1}))
 	require.NoError(t, mgr.savePersistedSnapshot(persistedJob{ID: "terminal", State: "completed", Seq: 2, Summary: "done"}))
@@ -219,7 +221,7 @@ func TestManagerShutdownFlushesPendingSnapshots(t *testing.T) {
 	mgr.persistDelay = time.Hour
 	require.NoError(t, mgr.savePersistedSnapshot(persistedJob{ID: "shutdown", State: "running", Seq: 7, LastMessage: "latest"}))
 
-	require.NoError(t, mgr.Shutdown(time.Second))
+	require.NoError(t, mgr.Shutdown(testwait.Timeout(time.Second)))
 	got, ok := readPersistedJob(filepath.Join(dir, "shutdown.json"))
 	require.True(t, ok)
 	assert.Equal(t, int64(7), got.Seq)

internal/jobs/testhelpers_test.go

@@ -197,8 +197,25 @@ func newTestManager(t *testing.T, prov provider.Provider, opts ...func(*Config))
 	}
 	mgr, _, err := NewManager(cfg)
 	require.NoError(t, err)
+	// Scaled, and reported rather than discarded.
+	//
+	// Shutdown returns when its timeout expires whether or not the workers
+	// stopped. Two seconds was ample on an idle machine and not ample under a
+	// parallel suite, and what happened then was not a slow test but a broken
+	// one: the workers were still running, still writing job records into the
+	// t.TempDir() Go was about to remove. On Windows a directory that is
+	// gaining files cannot be removed, so the test died in cleanup with
+	// "TempDir RemoveAll cleanup: ... The directory is not empty" -- naming
+	// neither the worker, nor the timeout, nor even the assertion it never
+	// reached. TestResubmit_IsOnlyAllowedOnce failed that way.
+	//
+	// The error says "N workers still running after ...", which is the one
+	// sentence that would have identified this immediately. Discarding it was
+	// how a shutdown that did not shut anything down looked like success.
 	t.Cleanup(func() {
-		_ = mgr.Shutdown(2 * time.Second)
+		if err := mgr.Shutdown(testwait.Timeout(2 * time.Second)); err != nil {
+			t.Errorf("jobs manager did not shut down cleanly: %v", err)
+		}
 	})
 	return mgr, tr
 }