refactor(agent): scope pending fallback to verification

Name the continuation fallback for its actual verification-only provenance so unrelated continuation paths cannot accidentally inherit its cron-delivery semantics.

agent/conversation_loop.py

@@ -613,11 +613,11 @@ def run_conversation(
     truncated_response_parts: List[str] = []
     compression_attempts = 0
     _turn_exit_reason = "unknown"  # Diagnostic: why the loop ended
-    # Last composed answer intentionally held back by an internal continuation
-    # gate.  If the continuation consumes the remaining budget, this is the
-    # best user-facing result available; it must not be confused with error or
+    # Last composed answer intentionally held back by a verification gate. If
+    # that continuation consumes the remaining budget, this is the best
+    # user-facing result available; it must not be confused with error or
     # recovery text produced by unrelated exit paths.
-    _pending_continuation_response = None
+    _pending_verification_response = None
 
     # Per-turn tally of consecutive successful credential-pool token refreshes,
     # keyed by (provider, pool-entry-id). A persistent upstream 401 lets
@@ -5182,7 +5182,7 @@ def run_conversation(
                     # continuation-budget exhaustion.  ``final_response`` itself
                     # must be cleared so the finalizer can distinguish this gate
                     # from unrelated error/recovery exits. (#61631)
-                    _pending_continuation_response = final_response
+                    _pending_verification_response = final_response
                     final_response = None
                     continue
 
@@ -5235,7 +5235,7 @@ def run_conversation(
                     agent._session_messages = messages
                     logger.debug("pre_verify nudge issued (attempt %d)",
                                  agent._pre_verify_nudges)
-                    _pending_continuation_response = final_response
+                    _pending_verification_response = final_response
                     final_response = None
                     continue
 
@@ -5321,7 +5321,7 @@ def run_conversation(
         original_user_message=original_user_message,
         _should_review_memory=_should_review_memory,
         _turn_exit_reason=_turn_exit_reason,
-        _pending_continuation_response=_pending_continuation_response,
+        _pending_verification_response=_pending_verification_response,
     )
 
 

agent/turn_finalizer.py

@@ -42,7 +42,7 @@ def finalize_turn(
     original_user_message,
     _should_review_memory,
     _turn_exit_reason,
-    _pending_continuation_response=None,
+    _pending_verification_response=None,
 ):
     """Run the post-loop finalization and return the turn ``result`` dict.
 
@@ -57,7 +57,7 @@ def finalize_turn(
     )
     continuation_budget_exhausted = (
         final_response is None
-        and bool(_pending_continuation_response)
+        and bool(_pending_verification_response)
         and budget_exhausted
     )
 
@@ -68,7 +68,7 @@ def finalize_turn(
         # one. Preserve that exact answer instead of replacing it with another
         # fallible model call. The explicit pending value is the provenance
         # guard: unrelated error/recovery exits can never enter this branch.
-        final_response = _pending_continuation_response
+        final_response = _pending_verification_response
         _turn_exit_reason = f"max_iterations_reached({api_call_count}/{agent.max_iterations})"
         iteration_limit_fallback = True
     elif final_response is None and budget_exhausted:

tests/agent/test_turn_finalizer_iteration_limit_exit.py

@@ -84,7 +84,7 @@ def _finalize(
     final_response,
     exit_reason,
     api_call_count=60,
-    pending_continuation_response=None,
+    pending_verification_response=None,
 ):
     return finalize_turn(
         agent,
@@ -100,7 +100,7 @@ def _finalize(
         original_user_message="task",
         _should_review_memory=False,
         _turn_exit_reason=exit_reason,
-        _pending_continuation_response=pending_continuation_response,
+        _pending_verification_response=pending_verification_response,
     )
 
 
@@ -114,7 +114,7 @@ def test_pending_verify_response_is_preserved_for_cron_delivery(monkeypatch):
         agent,
         final_response=None,
         exit_reason="unknown",
-        pending_continuation_response=report,
+        pending_verification_response=report,
     )
 
     assert result["final_response"] == report
@@ -131,7 +131,7 @@ def test_pending_pre_verify_response_is_preserved_on_budget_exhaustion(monkeypat
         agent,
         final_response=None,
         exit_reason="budget_exhausted",
-        pending_continuation_response=report,
+        pending_verification_response=report,
     )
 
     assert result["final_response"] == report
@@ -193,7 +193,7 @@ def test_pending_response_records_kanban_timeout(monkeypatch):
         agent,
         final_response=None,
         exit_reason="unknown",
-        pending_continuation_response="composed report",
+        pending_verification_response="composed report",
     )
 
     assert result["turn_exit_reason"] == "max_iterations_reached(60/60)"