From 32362ffbd9d7d4d0d2abc413e1dc8843f40c4ba7 Mon Sep 17 00:00:00 2001
From: Nick Sweeting <git@sweeting.me>
Date: Fri, 28 Aug 2026 11:27:46 -0700
Subject: [PATCH] Ignore historical hooks during plugin recovery

---
 archivebox/core/recovery_util.py |  3 +-
 archivebox/tests/test_cli_run.py | 79 ++++++++++++++++++++++++++++++++
 2 files changed, 80 insertions(+), 2 deletions(-)

diff --git a/archivebox/core/recovery_util.py b/archivebox/core/recovery_util.py
index ac702348..7837beaf 100644
--- a/archivebox/core/recovery_util.py
+++ b/archivebox/core/recovery_util.py
@@ -195,8 +195,7 @@ def recover_orchestrator_state(*, include_chrome: bool = False, crawl_id: str |
                 "status": ArchiveResult.StatusChoices.QUEUED,
             },
         )
-        process_is_newer = bool(process.started_at and (result.start_ts is None or process.started_at >= result.start_ts))
-        if result.status == ArchiveResult.StatusChoices.QUEUED or process_is_newer:
+        if created or result.status == ArchiveResult.StatusChoices.QUEUED:
             requeue_snapshot = False
             # A runner can die after the hook Process exits but before the
             # ProcessCompletedEvent projector links/finalizes ArchiveResult.
diff --git a/archivebox/tests/test_cli_run.py b/archivebox/tests/test_cli_run.py
index fc9a4325..79a17a86 100644
--- a/archivebox/tests/test_cli_run.py
+++ b/archivebox/tests/test_cli_run.py
@@ -2362,6 +2362,85 @@ class TestRecoverOrchestratorStateRedFailureModes:
         assert result.start_ts == newer_start
         assert result.process.started_at == newer_start
 
+    def test_recovery_does_not_replace_final_plugin_result_with_older_unlinked_hook(self):
+        import json
+        from datetime import timedelta
+
+        from django.utils import timezone
+
+        from archivebox.base_models.models import get_or_create_system_user_pk
+        from archivebox.core.models import ArchiveResult, Snapshot
+        from archivebox.core.recovery_util import recover_orchestrator_state
+        from archivebox.crawls.models import Crawl
+        from archivebox.machine.models import Machine, NetworkInterface, Process
+
+        crawl = Crawl.objects.create(
+            urls="https://example.com",
+            created_by_id=get_or_create_system_user_pk(),
+            status=Crawl.StatusChoices.SEALED,
+            retry_at=None,
+        )
+        snapshot = Snapshot.objects.create(url="https://example.com", crawl=crawl, status=Snapshot.StatusChoices.SEALED, retry_at=None)
+        machine = Machine.current(refresh=True)
+        iface = NetworkInterface.current(refresh=True)
+        start_ts = timezone.now() - timedelta(minutes=2)
+        stop_ts = timezone.now() - timedelta(minutes=1)
+        start_process = Process.objects.create(
+            machine=machine,
+            iface=iface,
+            process_type=Process.TypeChoices.HOOK,
+            worker_type="archiveresult",
+            pwd=str(snapshot.output_dir / "archivewebpage"),
+            cmd=["on_Snapshot__16_archivewebpage_start.js"],
+            status=Process.StatusChoices.EXITED,
+            retry_at=None,
+            exit_code=0,
+            started_at=start_ts,
+            ended_at=start_ts + timedelta(seconds=1),
+            stdout=json.dumps(
+                {
+                    "type": "ArchiveResult",
+                    "plugin": "archivewebpage",
+                    "hook_name": "on_Snapshot__16_archivewebpage_start",
+                    "status": "succeeded",
+                    "output_str": "recording started",
+                },
+            ),
+        )
+        stop_process = Process.objects.create(
+            machine=machine,
+            iface=iface,
+            process_type=Process.TypeChoices.HOOK,
+            worker_type="archiveresult",
+            pwd=str(snapshot.output_dir / "archivewebpage"),
+            cmd=["on_Snapshot__65_archivewebpage_stop.js"],
+            status=Process.StatusChoices.EXITED,
+            retry_at=None,
+            exit_code=0,
+            started_at=stop_ts,
+            ended_at=stop_ts + timedelta(seconds=1),
+        )
+        result = ArchiveResult.objects.create(
+            snapshot=snapshot,
+            plugin="archivewebpage",
+            hook_name="on_Snapshot__65_archivewebpage_stop",
+            status=ArchiveResult.StatusChoices.SUCCEEDED,
+            output_str="archivewebpage/archivewebpage.wacz",
+            start_ts=start_ts,
+            end_ts=stop_ts + timedelta(seconds=1),
+            process=stop_process,
+        )
+
+        recover_orchestrator_state()
+
+        result.refresh_from_db()
+        assert ArchiveResult.objects.filter(snapshot=snapshot, plugin="archivewebpage").count() == 1
+        assert result.hook_name == "on_Snapshot__65_archivewebpage_stop"
+        assert result.status == ArchiveResult.StatusChoices.SUCCEEDED
+        assert result.output_str == "archivewebpage/archivewebpage.wacz"
+        assert result.process == stop_process
+        assert not ArchiveResult.objects.filter(process=start_process).exists()
+
     def test_recovery_does_not_seal_queued_snapshot_waiting_for_future_retry_even_with_final_results(self):
         from datetime import timedelta