Fix TUI stuck in Working state after failed sub-agent transfer_task

When a child stream emitted an ErrorEvent, runSubSessionForwarding returned
immediately without reading remaining events. This caused StreamStoppedEvent
to never reach the TUI, leaking the streamDepth counter and leaving the UI
permanently stuck in the Working state.

Drain all remaining child events (including StreamStoppedEvent) before
returning on error, keeping streamDepth balanced.

Fixes #2255

Assisted-By: docker-agent

Author: dgageot

go.mod

@@ -185,7 +185,7 @@ require (
 	github.com/oasdiff/yaml3 v0.0.0-20250309153720-d2182401db90 // indirect
 	github.com/opencontainers/go-digest v1.0.0 // indirect
 	github.com/opencontainers/image-spec v1.1.1 // indirect
-	github.com/patrickmn/go-cache v2.1.0+incompatible // indirect
+	github.com/patrickmn/go-cache v2.1.0+incompatible
 	github.com/perimeterx/marshmallow v1.1.5 // indirect
 	github.com/pjbgf/sha1cd v0.3.2 // indirect
 	github.com/pmezard/go-difflib v1.0.0 // indirect

pkg/runtime/agent_delegation.go

@@ -125,9 +125,15 @@ func newSubSession(parent *session.Session, cfg SubSessionConfig, childAgent *ag
 // This is the "interactive" path used by transfer_task where the parent agent
 // loop is blocked while the child executes.
 func (r *LocalRuntime) runSubSessionForwarding(ctx context.Context, parent, child *session.Session, span trace.Span, evts chan Event, callerAgent string) (*tools.ToolCallResult, error) {
-	for event := range r.RunStream(ctx, child) {
+	childEvents := r.RunStream(ctx, child)
+	for event := range childEvents {
 		evts <- event
 		if errEvent, ok := event.(*ErrorEvent); ok {
+			// Drain remaining events (including StreamStoppedEvent) so the
+			// TUI's streamDepth counter stays balanced.
+			for remaining := range childEvents {
+				evts <- remaining
+			}
 			span.RecordError(fmt.Errorf("%s", errEvent.Error))
 			span.SetStatus(codes.Error, "sub-session error")
 			return nil, fmt.Errorf("%s", errEvent.Error)