feat(lease): make the author task heartbeat load-bearing (#790 Slice A)

Slice A of Issue #790, per the controller reassessment in comment 13958. Does
not close the issue: terminal retirement (Slice B) and the read-side generation
check plus the #760 renewal re-scope (Slice C) are deliberately not implemented.

The defect. `issue_lock_store.assess_lock_freshness` parsed `last_heartbeat_at`
and then never consulted it. Liveness was decided by an absolute four-hour
`expires_at` and by PID liveness, and the recorded PID is the long-lived MCP
daemon rather than the authoring task, so an abandoned claim stayed live for the
full four hours. A tree-wide search found the field written in exactly one place
and advanced by nothing. Issue #787 / PR #789 hit this; Issue #760 / PR #791 hit
it again, blocking reconciliation for over five hours after its work had landed.

A1 — central policy. New `lease_policy` declares every duration for every task
class in one place: author initial/sliding TTL 10 minutes, heartbeat cadence 2,
stale warning 5, missed-heartbeat grace 10, absolute cap 8 hours, recovery grace
10, terminal race-drain 2. It ships first so the first heartbeat and TTL
behavior to run reads from it (AC-N7). The duplicated four-hour literal is gone
from both `issue_lock_store` and `gitea_mcp_server`. Reviewer, merger, and
conflict-fix classes are declared but not rewired — Slice C moves those call
sites — and a test asserts the declaration still equals the constants #747 and
`pr_work_lease` own, so the two cannot drift apart unnoticed.

A2 — load-bearing freshness, with two deliberate asymmetries. An alive PID never
establishes freshness anywhere (AC-N2); it is recorded as evidence and no branch
returns live because of it. A dead PID still marks a lease stale, and that band
still precedes every heartbeat evaluation, so #753 dead-session recovery keys on
exactly the classification it always did. New bands `stale_missed_heartbeat` and
`stale_absolute_cap` are classified in `branch_cleanup_guard` rather than
falling through to unknown-status, and still block unless the ownership record
proves `reclaim_allowed is True`. A heartbeat lease carrying no heartbeat is
contradictory and fails closed. `assess_expired_lock_reclaim` accepts a lapsed
heartbeat as reclaim grounds for heartbeat-lifecycle leases only: under this
lifecycle the heartbeat is the liveness proof, and also requiring a dead PID
would reinstate the original defect.

A3/A4 — task-session identity and the writer. `mint_task_session_id` produces an
ownership key containing no process identifier, since the daemon PID is reused
by every task it serves and identifies none of them. `heartbeat_session_lock`
writes inside the existing per-issue flock under the #772 generation
compare-and-swap, verifying exact issue, branch, realpath-normalized worktree,
claimant username, claimant profile, and recorded session identifier. It cannot
acquire, take over, or revive: a lease past its grace is refused and must use
the reclaim path, so a session that stopped proving liveness cannot restore
ownership retroactively. New `gitea_heartbeat_issue_lock` gates on the same
authority as `lock_issue`, being strictly narrower.

A5 — legacy compatibility (AC-N8). The explicit `lifecycle_version` marker, never
a timestamp comparison, discriminates legacy from heartbeat leases: a legacy lock
has `last_heartbeat_at == created_at` forever precisely because nothing advanced
it, and a freshly minted heartbeat lease has them equal too, so the equality
carries no information in either direction. Legacy locks keep their recorded
absolute expiry and are never evaluated against the short grace, so deployment
cannot make an existing claim instantly reclaimable. They leave that state only
by terminal retirement (Slice B) or by `rebind_legacy_lock`, which re-verifies
the exact owner and mints a genuine identifier and first heartbeat while
preserving the original claim under `legacy_origin`. Rebinding a lapsed legacy
lease is refused; that belongs to #760 renewal or #601 reclaim.

A6 — native coverage. Review #499 proved assessor-level tests miss discard
points, so `tests/test_issue_790_heartbeat_mcp_path.py` drives the real tools
against a real git repository and a real durable lock: lock creation and
read-back, policy window, freshness, survival of `verify_lock_for_mutation`,
invariance of the duplicate-work and linked-open-PR gates, CAS rejection,
foreign-session and foreign-claimant refusal, alive-PID-only refusal, missed
heartbeat, legacy protection on deployment, and legacy rebinding.

Tests. New suites 55 passed. Lock and lease regression set (issue_lock_store,
lease_lifecycle, #753, #755, #760 x2, #768, #772, lock registration, worktree,
adoption, duplicate gate, branch cleanup guard, capability invariants, claim
heartbeat, worktrees) 383 passed with 98 subtests. Full suite 4295 passed, 11
failed, 6 skipped, 499 subtests passed, against a clean master baseline worktree
at 620ed6e9 that reports 11 failed and 4240 passed — the same eleven node IDs.
The 55-test delta is exactly the new suites; no new failures.

Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>
Claude-Session: https://claude.ai/code/session_011u6GKSJwwrrYjguPjs1aK5

gitea_mcp_server.py

@@ -2007,6 +2007,7 @@ import allocator_dependencies  # noqa: E402
 import dependency_graph  # noqa: E402  # #784 durable dependency edges
 import control_plane_db  # noqa: E402
 import lease_lifecycle  # noqa: E402
+import lease_policy  # noqa: E402
 import workflow_dashboard  # noqa: E402  # #605 live queue/lease dashboard
 import incident_bridge  # noqa: E402
 import sentry_observability  # noqa: E402  (#606 optional Sentry observability)
@@ -2248,7 +2249,6 @@ import canonical_comment_validator as ccv  # noqa: E402
 # GITEA_ISSUE_LOCK_DIR, bound to the current MCP session via a per-PID pointer.
 # Legacy global path retained only for test/doc references — do not seed manually.
 ISSUE_LOCK_FILE = "/tmp/gitea_issue_lock.json"
-WORK_LEASE_TTL_HOURS = 4
 AUTHOR_ISSUE_WORK_LEASE = "author_issue_work"
 VALID_WORK_LEASE_OPERATIONS = frozenset({
     AUTHOR_ISSUE_WORK_LEASE,
@@ -2548,7 +2548,12 @@ def _build_author_issue_work_lease(
     host: str | None,
 ) -> dict:
     created = _work_lease_now()
-    expires = created + timedelta(hours=WORK_LEASE_TTL_HOURS)
+    # #790 Slice A: the window comes from the central policy, not a literal here.
+    # It is also now a *sliding* window — the lease lives ``initial_ttl_minutes``
+    # past its last valid heartbeat rather than a fixed four hours past its
+    # creation, so an abandoned task stops holding the claim within one TTL.
+    policy = lease_policy.policy_for(lease_policy.TASK_CLASS_AUTHOR_ISSUE_WORK)
+    expires = created + timedelta(minutes=policy.initial_ttl_minutes)
     return {
         "operation_type": AUTHOR_ISSUE_WORK_LEASE,
         "issue_number": issue_number,
@@ -2559,6 +2564,15 @@ def _build_author_issue_work_lease(
         "created_at": _work_lease_timestamp(created),
         "expires_at": _work_lease_timestamp(expires),
         "last_heartbeat_at": _work_lease_timestamp(created),
+        # #790 AC-N1: the ownership key for this task. Distinct from the recorded
+        # PID, which is the shared daemon and identifies no individual task.
+        "task_session_id": issue_lock_store.mint_task_session_id(
+            AUTHOR_ISSUE_WORK_LEASE
+        ),
+        # #790 AC-N8: the explicit lifecycle marker. Its absence — never a
+        # timestamp comparison — is what makes a lock legacy.
+        "lifecycle_version": lease_policy.LIFECYCLE_HEARTBEAT_V1,
+        "heartbeat_count": 1,
     }
 
 
@@ -4328,6 +4342,138 @@ def gitea_lock_issue(
     return result
 
 
+@mcp.tool()
+def gitea_heartbeat_issue_lock(
+    issue_number: int,
+    branch_name: str,
+    task_session_id: str | None = None,
+    remote: str = "dadeschools",
+    host: str | None = None,
+    org: str | None = None,
+    repo: str | None = None,
+    worktree_path: str | None = None,
+    expected_generation: int | None = None,
+) -> dict:
+    """Prove an owned author issue lease is still active (#790 Slice A).
+
+    The task-liveness signal the lifecycle was missing. Before this, an author
+    lease carried a fixed four-hour expiry that nothing could shorten, and the
+    only liveness evidence was the recorded PID — the long-lived MCP daemon,
+    which stays alive across every task it serves and so proved nothing about
+    whether the authoring task still held the work.
+
+    Each successful call slides the lease ``initial_ttl_minutes`` past *now*
+    from the central policy, so an actively heartbeating session is never
+    evicted while an abandoned one releases its claim within one TTL.
+
+    What this tool cannot do, by construction:
+
+    * **Acquire.** It refuses when no durable lock exists.
+    * **Take over.** Exact issue, branch, realpath-normalized worktree,
+      claimant username, claimant profile, and recorded task-session identifier
+      must all match; a superseded session holding an older identifier is
+      refused.
+    * **Revive.** A lease already past its grace is not heartbeatable — that
+      would let a session restore ownership it had stopped proving. It must use
+      the sanctioned reclaim path, which mints a new generation.
+
+    A lock predating the heartbeat lifecycle is rebound rather than heartbeated:
+    its exact owner is re-verified and a genuine task-session identifier and
+    first heartbeat are minted (#790 AC-N8). The rebind is decided server-side
+    from the durable lifecycle marker; there is no caller-facing switch.
+
+    Args:
+        issue_number: The locked issue number.
+        branch_name: The branch recorded on the lock.
+        task_session_id: The identifier this session received when it acquired
+            or rebound the lock. It is a fencing token, not an ownership
+            assertion: it is compared against durable state and can only ever
+            cause a refusal, never grant anything. Omitted only when rebinding a
+            legacy lock, which has no identifier yet and mints one.
+        remote: Known instance — 'dadeschools' or 'prgs'.
+        host: Override the Gitea host.
+        org: Override the owner/organization.
+        repo: Override the repository name.
+        worktree_path: Author worktree recorded on the lock.
+        expected_generation: Optional fencing value. The per-issue flock already
+            serializes the read and the write, so this is for a caller that
+            wants to pin the generation it last observed across calls; a moved
+            generation fails closed.
+
+    Returns:
+        dict with 'success', 'performed', the sliding 'expires_at',
+        'last_heartbeat_at', 'lock_generation', 'task_session_id', the applied
+        'policy', and post-write 'freshness'; on refusal 'success'/'performed'
+        False with 'reasons' naming exactly what did not match.
+    """
+    blocked = _profile_permission_block(
+        task_capability_map.required_permission("heartbeat_issue_lock"),
+        issue_number=issue_number,
+        remote=remote,
+        host=host,
+        org=org,
+        repo=repo,
+        org_explicit=org is not None,
+        repo_explicit=repo is not None,
+    )
+    if blocked:
+        return blocked
+
+    resolved_worktree = issue_lock_worktree.resolve_author_worktree_path(
+        worktree_path, _canonical_local_git_root()
+    )
+    h, o, r = _resolve(remote, host, org, repo)
+    claimant = _work_lease_claimant(h)
+    identity = claimant.get("username")
+    profile = claimant.get("profile")
+
+    existing = _load_existing_issue_lock(
+        remote=remote, org=o, repo=r, issue_number=issue_number
+    )
+    if not existing:
+        return {
+            "success": False,
+            "performed": False,
+            "issue_number": issue_number,
+            "reasons": [
+                f"no durable lock for issue #{issue_number}; heartbeat cannot "
+                "acquire a claim (fail closed)"
+            ],
+        }
+
+    if issue_lock_store.is_legacy_lease(existing):
+        # AC-N8 exit route one: canonical exact-owner rebinding. The other exit
+        # is terminal retirement, which is Slice B.
+        outcome = issue_lock_store.rebind_legacy_lock(
+            remote=remote,
+            org=o,
+            repo=r,
+            issue_number=issue_number,
+            branch_name=branch_name,
+            worktree_path=resolved_worktree,
+            identity=identity,
+            profile=profile,
+            expected_generation=expected_generation,
+        )
+        outcome["operation"] = "legacy_rebind"
+        return outcome
+
+    outcome = issue_lock_store.heartbeat_session_lock(
+        remote=remote,
+        org=o,
+        repo=r,
+        issue_number=issue_number,
+        branch_name=branch_name,
+        worktree_path=resolved_worktree,
+        identity=identity,
+        profile=profile,
+        task_session_id=str(task_session_id or ""),
+        expected_generation=expected_generation,
+    )
+    outcome["operation"] = "heartbeat"
+    return outcome
+
+
 @mcp.tool()
 def gitea_assess_work_issue_duplicate(
     issue_number: int,