Stopping a live flow ends the calls it has running, on workers and in the local pool, and deleting one now ends its local calls too; what a call returns after its flow was stopped or deleted is dropped by a per-flow count the node compares, so a flow started again never takes an old call's answer. A live remote call's id carries a suffix, so two calls of one node no longer share it, and a call is registered before it queues for a slot, so ending its flow or run reaches it and it is never sent; a run's cancel goes through the call's own thread as well

backend/tests/flow/test_service_node.py

@@ -180,10 +180,11 @@ def test_a_service_that_never_gets_healthy_reads_down(monkeypatch):
     asyncio.run(scenario())
 
 
-def test_a_stop_right_after_the_start_ends_the_serve_after_it(monkeypatch):
+def test_a_stop_right_after_the_start_leaves_no_server_running(monkeypatch):
     """The stop's cancel went out on a thread of its own, so with the serve's
     thread slow to start it reached the worker first. A worker ignores a cancel
     for a call it does not know, and ran that server until its socket dropped.
+    Ended before its thread has sent it, the serve now never goes out.
     """
     monkeypatch.setattr(remote, "SILENCE_S", 1.0)
     monkeypatch.setattr(service, "STOP_WAIT_S", 1.0)
@@ -228,4 +229,4 @@ def test_a_stop_right_after_the_start_ends_the_serve_after_it(monkeypatch):
         await node.stop()
 
     asyncio.run(scenario())
-    assert [frame["op"] for frame in agent.sent] == ["serve", "cancel"]
+    assert agent.sent == []