wip(k1): checkpoint connection recovery rewrite

Capture the current unreleased K1 connection, recovery, lifecycle, viewer, and test work as a single known-bad baseline for subsequent fixes.

tests/test_active_session_lease.py

@@ -1,9 +1,13 @@
 from __future__ import annotations
 
+import os
 from pathlib import Path
 
+import pytest
+
 from k1link.device_plugins.xgrids_k1.archive import discover_legacy_viewer_sessions
 from k1link.sessions import ActiveSessionLease, recover_stale_active_session_marker
+from k1link.sessions import active as active_session_module
 
 
 def test_active_session_lease_hides_live_evidence_until_release(tmp_path: Path) -> None:
@@ -38,3 +42,65 @@ def test_startup_recovery_removes_only_an_unlocked_stale_marker(tmp_path: Path)
     assert recover_stale_active_session_marker(sessions) is True
     assert not marker.exists()
     assert recover_stale_active_session_marker(sessions) is False
+
+
+def test_active_session_release_retries_unlink_without_dropping_lock(
+    tmp_path: Path,
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    sessions = tmp_path / "sessions"
+    session = sessions / "20260812T231000Z_viewer_live"
+    lease = ActiveSessionLease.acquire(sessions, session)
+    marker = sessions / ".current_session"
+    original_unlink = Path.unlink
+    attempts = 0
+
+    def fail_once(path: Path, *args: object, **kwargs: object) -> None:
+        nonlocal attempts
+        if path == marker and attempts == 0:
+            attempts += 1
+            raise OSError("injected marker unlink failure")
+        original_unlink(path, *args, **kwargs)
+
+    monkeypatch.setattr(Path, "unlink", fail_once)
+    with pytest.raises(OSError, match="injected marker unlink failure"):
+        lease.release()
+
+    assert marker.exists()
+    assert lease._released is False  # noqa: SLF001
+    os.fstat(lease._descriptor)  # noqa: SLF001
+    assert recover_stale_active_session_marker(sessions) is False
+
+    lease.release()
+    assert lease._released is True  # noqa: SLF001
+    assert not marker.exists()
+
+
+def test_active_session_release_retries_directory_fsync_after_unlink(
+    tmp_path: Path,
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    sessions = tmp_path / "sessions"
+    session = sessions / "20260812T231100Z_viewer_live"
+    lease = ActiveSessionLease.acquire(sessions, session)
+    original_fsync = active_session_module._fsync_directory_strict
+    attempts = 0
+
+    def fail_once(path: Path) -> None:
+        nonlocal attempts
+        if attempts == 0:
+            attempts += 1
+            raise OSError("injected directory fsync failure")
+        original_fsync(path)
+
+    monkeypatch.setattr(active_session_module, "_fsync_directory_strict", fail_once)
+    with pytest.raises(OSError, match="injected directory fsync failure"):
+        lease.release()
+
+    assert lease._marker_removed is True  # noqa: SLF001
+    assert lease._released is False  # noqa: SLF001
+    os.fstat(lease._descriptor)  # noqa: SLF001
+
+    lease.release()
+    lease.release()
+    assert lease._released is True  # noqa: SLF001