feat(agent-call): implement remediation plan

tests/test_agent_call.py

@@ -4,6 +4,7 @@ import copy
 import json
 import tempfile
 import unittest
+from datetime import timedelta
 from pathlib import Path
 
 from agent_call.core import (
@@ -104,6 +105,53 @@ class AgentCallTests(unittest.TestCase):
         attempts = self.service.store.one("SELECT COUNT(*) AS n FROM attempts")["n"]
         self.assertEqual(calls, 1)
         self.assertEqual(attempts, 1)
+        self.assertEqual(
+            self.service.store.one(
+                "SELECT status FROM executions WHERE tenant_id=? AND execution_id=?",
+                ("tenant-demo", "exec_once"),
+            )["status"],
+            "completed",
+        )
+
+    def test_unhealthy_cell_isolated_without_blocking_healthy_capacity(self) -> None:
+        self.service.set_cell_health("cell-a", False)
+        self.assertTrue(self.service.ready())
+        snapshot = self.run_command(self.command("cmd_healthy_cell"))
+        call = self.service.get_call("tenant-demo", snapshot["call_id"])
+        self.assertEqual(call["call_state"], "ended")
+        self.assertEqual(
+            self.service.store.one(
+                "SELECT cell_id FROM calls WHERE call_id=?", (snapshot["call_id"],)
+            )["cell_id"],
+            "cell-b",
+        )
+
+    def test_missing_tenant_route_closes_readiness(self) -> None:
+        self.service._unrouteable_tenants.add("tenant-demo")
+        snapshot = self.service.readiness()
+        self.assertFalse(snapshot["checks"]["routes"])
+        self.assertFalse(snapshot["ready"])
+
+    def test_expired_scheduler_lease_fences_old_instance(self) -> None:
+        body = self.command("cmd_lease_fence")
+        self.service.publish_execute(body)
+        self.service.ingest_commands()
+        second = AgentCallService(
+            db_path=Path(self.temp.name) / "executor.sqlite3",
+            object_dir=Path(self.temp.name) / "objects-second",
+            broker=self.broker,
+        )
+        try:
+            with self.service.store.tx() as con:
+                con.execute(
+                    "UPDATE leases SET expires_at=? WHERE lease_name='scheduler'",
+                    (iso(utcnow() - timedelta(seconds=1)),),
+                )
+            self.assertEqual(second.schedule_once(), 1)
+            self.assertEqual(self.service.schedule_once(), 0)
+            second.wait_for_idle()
+        finally:
+            second.stop()
 
     def test_controls_use_cas_and_old_commands_do_not_resume(self) -> None:
         pause = self.service.control_task(