close worktree transport and lifecycle contract gaps

internal/orchestrator/orchestrator.go

@@ -7,6 +7,7 @@ import (
 	"context"
 	"encoding/json"
 	"fmt"
+	"orchestra/internal/continuity"
 	"orchestra/internal/domain"
 	"orchestra/internal/herdr"
 	"orchestra/internal/store"
@@ -28,8 +29,9 @@ type Adapters interface {
 // to be a clone containing the project's remote; callers may set a separate
 // root per deployment.
 type GitWorktrees struct {
-	Root string
-	Repo string
+	Root        string
+	Repo        string
+	TaskFileSHA string
 }
 
 func (w GitWorktrees) Create(ctx context.Context, t domain.Task) (string, error) {
@@ -41,6 +43,11 @@ func (w GitWorktrees) Create(ctx context.Context, t domain.Task) (string, error)
 	}
 	p := filepath.Join(w.Root, t.ID)
 	if _, err := os.Stat(p); err == nil {
+		if w.TaskFileSHA != "" {
+			if err := continuity.VerifyTaskFile(p, w.TaskFileSHA); err != nil {
+				return "", err
+			}
+		}
 		return p, nil
 	}
 	branch := "orchestra/" + t.ID
@@ -48,6 +55,11 @@ func (w GitWorktrees) Create(ctx context.Context, t domain.Task) (string, error)
 	if out, err := cmd.CombinedOutput(); err != nil {
 		return "", fmt.Errorf("%s: %w", string(out), err)
 	}
+	if w.TaskFileSHA != "" {
+		if err := continuity.VerifyTaskFile(p, w.TaskFileSHA); err != nil {
+			return "", err
+		}
+	}
 	return p, nil
 }
 
@@ -69,6 +81,33 @@ type Coordinator struct {
 	mu        sync.Mutex
 	sessions  map[string]herdr.Session
 	loaded    bool
+	healthMu  sync.RWMutex
+	health    MonitorHealth
+}
+
+type MonitorHealth struct {
+	Running   bool      `json:"running"`
+	LastRun   time.Time `json:"last_run"`
+	LastError string    `json:"last_error,omitempty"`
+	Expired   int       `json:"expired"`
+}
+
+func (c *Coordinator) MonitorHealth() MonitorHealth {
+	c.healthMu.RLock()
+	defer c.healthMu.RUnlock()
+	return c.health
+}
+func (c *Coordinator) setMonitorHealth(err error, expired int) {
+	c.healthMu.Lock()
+	defer c.healthMu.Unlock()
+	c.health.Running = err == nil
+	c.health.LastRun = time.Now().UTC()
+	c.health.Expired += expired
+	if err != nil {
+		c.health.LastError = err.Error()
+	} else {
+		c.health.LastError = ""
+	}
 }
 
 func (c *Coordinator) loadSessions() {
@@ -128,6 +167,7 @@ func (c *Coordinator) Reconcile(ctx context.Context) error {
 // and frees the lease for pickup by the router.
 func (c *Coordinator) Monitor(ctx context.Context, hard float64, interval time.Duration) error {
 	if err := c.Reconcile(ctx); err != nil {
+		c.setMonitorHealth(err, 0)
 		return err
 	}
 	if interval <= 0 {
@@ -138,12 +178,58 @@ func (c *Coordinator) Monitor(ctx context.Context, hard float64, interval time.D
 	for {
 		select {
 		case <-ctx.Done():
+			c.healthMu.Lock()
+			c.health.Running = false
+			c.healthMu.Unlock()
 			return ctx.Err()
 		case <-t.C:
+			expired, err := c.expire(ctx)
+			c.setMonitorHealth(err, len(expired))
+			if err != nil {
+				continue
+			}
 			c.rotate(ctx, hard)
 		}
 	}
 }
+
+func (c *Coordinator) expire(ctx context.Context) ([]domain.Event, error) {
+	// pane.exited is the low-latency path; lease expiry below remains the
+	// authoritative backstop when herdr misses an exit notification.
+	c.loadSessions()
+	c.mu.Lock()
+	for taskID, s := range c.sessions {
+		if t, ok := c.Store.Task(taskID); ok && t.State == domain.StateLeased {
+			if a, ae := c.Adapters.Adapter(s.Harness); ae == nil {
+				if p, ok := a.(herdr.PaneExit); ok {
+					if exited, ee := p.PaneExited(ctx, s); ee == nil && exited {
+						b, _ := json.Marshal(map[string]string{"reason": "pane_exited", "harness_id": s.Harness})
+						_ = c.Store.Append(domain.Event{ID: domain.NewID(), Type: "TaskReleased", TaskID: taskID, Version: t.Version + 1, Payload: b})
+					}
+				}
+			}
+		}
+	}
+	c.mu.Unlock()
+	events, err := c.Store.ExpireLeases(time.Now())
+	if err != nil {
+		return events, err
+	}
+	for _, e := range events {
+		c.loadSessions()
+		c.mu.Lock()
+		s, ok := c.sessions[e.TaskID]
+		delete(c.sessions, e.TaskID)
+		if ok {
+			if a, ae := c.Adapters.Adapter(s.Harness); ae == nil {
+				_ = a.Kill(ctx, s)
+			}
+		}
+		_ = c.saveSessionsLocked()
+		c.mu.Unlock()
+	}
+	return events, nil
+}
 func (c *Coordinator) rotate(ctx context.Context, hard float64) {
 	c.loadSessions()
 	c.mu.Lock()