fix: recover from phantom 'running' state after crash/restart

The run-cancel registry is in-memory; a container restart mid-run leaves
accounts/tasks persisted as 'running' with no goroutine, wedging cancel
(not-in-map -> 409) and blocking remove/re-run.

- startup: ResetRunningOnStartup clears stale 'running' -> 'idle' on boot
- cancel handler: when no live goroutine, ClearStuckAccount + ReconcileTaskStatus
  reset the stuck account (and its task) instead of returning 409

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MMHQTtnQtQqL8muAXHr9kd

internal/store/recovery.go

@@ -0,0 +1,44 @@
+package store
+
+import "context"
+
+// ResetRunningOnStartup clears phantom "running" statuses left behind when the
+// process died mid-run (crash, container restart). A fresh process has no
+// in-flight goroutines, so any persisted "running" is stale and would otherwise
+// wedge the task (the run-guard refuses to start, and accounts can't be edited).
+// Returns how many task and account rows were reset.
+func (s *Store) ResetRunningOnStartup(ctx context.Context) (tasks int64, accounts int64, err error) {
+	ct, err := s.Pool.Exec(ctx, `UPDATE accounts SET status='idle' WHERE status='running'`)
+	if err != nil {
+		return 0, 0, err
+	}
+	accounts = ct.RowsAffected()
+	ct, err = s.Pool.Exec(ctx, `UPDATE tasks SET status='idle' WHERE status='running'`)
+	if err != nil {
+		return 0, accounts, err
+	}
+	tasks = ct.RowsAffected()
+	return tasks, accounts, nil
+}
+
+// ClearStuckAccount resets an account stuck in "running" (no live goroutine) to
+// "idle". Returns true if a stuck row was actually cleared.
+func (s *Store) ClearStuckAccount(ctx context.Context, accountID int64) (bool, error) {
+	ct, err := s.Pool.Exec(ctx, `UPDATE accounts SET status='idle' WHERE id=$1 AND status='running'`, accountID)
+	if err != nil {
+		return false, err
+	}
+	return ct.RowsAffected() == 1, nil
+}
+
+// ReconcileTaskStatus moves a task out of "running" once none of its accounts
+// are still running — used after clearing a stuck account so the task can be
+// re-run or deleted again.
+func (s *Store) ReconcileTaskStatus(ctx context.Context, taskID int64) error {
+	_, err := s.Pool.Exec(ctx,
+		`UPDATE tasks SET status='idle'
+		 WHERE id=$1 AND status='running'
+		   AND NOT EXISTS (SELECT 1 FROM accounts WHERE task_id=$1 AND status='running')`,
+		taskID)
+	return err
+}

internal/store/recovery_test.go

@@ -0,0 +1,62 @@
+package store
+
+import (
+	"context"
+	"testing"
+)
+
+func TestResetRunningOnStartup(t *testing.T) {
+	s := testStore(t)
+	ctx := context.Background()
+	e1, _ := s.CreateEndpoint(ctx, Endpoint{RoleLabel: "s", Host: "a", Port: 993, TLSMode: "ssl"})
+	e2, _ := s.CreateEndpoint(ctx, Endpoint{RoleLabel: "d", Host: "b", Port: 993, TLSMode: "ssl"})
+	taskID, _ := s.CreateTask(ctx, Task{Name: "t", SrcEndpointID: e1, DstEndpointID: e2})
+	accID, _ := s.CreateAccount(ctx, Account{TaskID: taskID, SrcLogin: "u", SrcPassEnc: "x", DstLogin: "v", DstPassEnc: "y"})
+	// simulate a crash mid-run
+	_ = s.SetTaskStatus(ctx, taskID, "running")
+	_ = s.SetAccountStatus(ctx, accID, "running")
+
+	tn, an, err := s.ResetRunningOnStartup(ctx)
+	if err != nil {
+		t.Fatalf("reset: %v", err)
+	}
+	if tn != 1 || an != 1 {
+		t.Fatalf("reset counts tasks=%d accounts=%d, want 1/1", tn, an)
+	}
+	task, _ := s.GetTask(ctx, taskID)
+	if task.Status == "running" {
+		t.Fatal("task still running after reset")
+	}
+	accs, _ := s.ListAccountsByTask(ctx, taskID)
+	if accs[0].Status == "running" {
+		t.Fatal("account still running after reset")
+	}
+}
+
+func TestClearStuckAccountAndReconcile(t *testing.T) {
+	s := testStore(t)
+	ctx := context.Background()
+	e1, _ := s.CreateEndpoint(ctx, Endpoint{RoleLabel: "s", Host: "a", Port: 993, TLSMode: "ssl"})
+	e2, _ := s.CreateEndpoint(ctx, Endpoint{RoleLabel: "d", Host: "b", Port: 993, TLSMode: "ssl"})
+	taskID, _ := s.CreateTask(ctx, Task{Name: "t", SrcEndpointID: e1, DstEndpointID: e2})
+	accID, _ := s.CreateAccount(ctx, Account{TaskID: taskID, SrcLogin: "u", SrcPassEnc: "x", DstLogin: "v", DstPassEnc: "y"})
+	_ = s.SetTaskStatus(ctx, taskID, "running")
+	_ = s.SetAccountStatus(ctx, accID, "running")
+
+	cleared, err := s.ClearStuckAccount(ctx, accID)
+	if err != nil || !cleared {
+		t.Fatalf("clear stuck: cleared=%v err=%v", cleared, err)
+	}
+	// second call finds nothing to clear
+	again, _ := s.ClearStuckAccount(ctx, accID)
+	if again {
+		t.Fatal("second clear should be false")
+	}
+	if err := s.ReconcileTaskStatus(ctx, taskID); err != nil {
+		t.Fatalf("reconcile: %v", err)
+	}
+	task, _ := s.GetTask(ctx, taskID)
+	if task.Status == "running" {
+		t.Fatal("task should no longer be running after reconcile")
+	}
+}