feat: cascade retry deps when running a task with failed dependencies

When /run is called on a CANCELLED/FAILED task that has deps in a terminal
failure state, automatically reset and resubmit those deps so the task
isn't immediately re-cancelled by the pool's dep check.

Also update reset-failed-tasks script to handle CANCELLED tasks and clean
up preserved sandbox workspaces.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

1 files changed, 39 insertions, 0 deletions

diff --git a/internal/api/server.go b/internal/api/server.go
index 38108de..8eba829 100644
--- a/internal/api/server.go
+++ b/internal/api/server.go
@@ -571,6 +571,41 @@ func (s *Server) handleGetTask(w http.ResponseWriter, r *http.Request) {
 	}
 	writeJSON(w, http.StatusOK, s.enrichTask(t))
 }
+// retryableDepStates are the states from which a dependency can be retried
+// when cascading a retry from a dependent task.
+var retryableDepStates = map[task.State]bool{
+	task.StateFailed:         true,
+	task.StateTimedOut:       true,
+	task.StateCancelled:      true,
+	task.StateBudgetExceeded: true,
+}
+
+// cascadeRetryDeps resets any dependency (recursively) that is in a retryable
+// terminal state, and submits it to the pool. This ensures that retrying a
+// CANCELLED task that was blocked by a failed dep will also restart that dep.
+func (s *Server) cascadeRetryDeps(ctx context.Context, t *task.Task) {
+	for _, depID := range t.DependsOn {
+		dep, err := s.store.GetTask(depID)
+		if err != nil {
+			s.logger.Warn("cascadeRetryDeps: dep not found", "depID", depID)
+			continue
+		}
+		if !retryableDepStates[dep.State] {
+			continue
+		}
+		// Recursively cascade first (depth-first so root deps go first).
+		s.cascadeRetryDeps(ctx, dep)
+		reset, err := s.store.ResetTaskForRetry(depID)
+		if err != nil {
+			s.logger.Warn("cascadeRetryDeps: reset failed", "depID", depID, "error", err)
+			continue
+		}
+		if submitErr := s.pool.Submit(ctx, reset); submitErr != nil {
+			s.logger.Warn("cascadeRetryDeps: submit failed", "depID", depID, "error", submitErr)
+		}
+	}
+}
+
 func (s *Server) handleRunTask(w http.ResponseWriter, r *http.Request) {
 	id := r.PathValue("id")
 	agentParam := r.URL.Query().Get("agent") // Use a different name to avoid confusion
@@ -619,6 +654,10 @@ func (s *Server) handleRunTask(w http.ResponseWriter, r *http.Request) {
 	}
 	// The task `t` now has the correct agent configuration.
 
+	// 6. Cascade-retry any deps that are in a terminal failure state so the
+	// task isn't immediately re-cancelled by checkDepsReady.
+	s.cascadeRetryDeps(r.Context(), originalTask)
+
 	if err := s.pool.Submit(context.Background(), t); err != nil {
 		writeJSON(w, http.StatusServiceUnavailable, map[string]string{"error": fmt.Sprintf("executor pool: %v", err)})
 		return