Keep failed cumulative batches atomic

README.md

@@ -11,11 +11,11 @@ integration PRs, follows `main` through production, and pauses after failures.
 
 ## Install
 
-Install the reviewed `v0.2.23` source commit directly from GitHub:
+Install the reviewed `v0.2.24` source commit directly from GitHub:
 
 ```bash
 python3 -m pip install \
-  'deploybot-merge-queue[mcp] @ git+https://github.com/Forward-Future/DeployBot.git@de0819770c2496b5048488c2a6a207be0378af06'
+  'deploybot-merge-queue[mcp] @ git+https://github.com/Forward-Future/DeployBot.git@73004ea7c9dcb81e7f1281c0687aea0897d1571d'
 deploybot init
 ```
 
@@ -95,7 +95,7 @@ worker can dispatch deployment when GitHub suppresses the `workflow_run` event
 for token-dispatched CI. Pin the Action to the full reviewed release commit:
 
 ```yaml
-- uses: Forward-Future/DeployBot@de0819770c2496b5048488c2a6a207be0378af06
+- uses: Forward-Future/DeployBot@73004ea7c9dcb81e7f1281c0687aea0897d1571d
 ```
 
 The Action uses GitHub's built-in workflow token. GitHub intentionally does not

adapters/claude-code/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "deploybot",
-  "version": "0.2.23",
+  "version": "0.2.24",
   "description": "DeployBot: a provider-neutral GitHub merge queue for coding agents",
   "author": {
     "name": "DeployBot contributors"

adapters/claude-code/.mcp.json

@@ -4,7 +4,7 @@
       "command": "uvx",
       "args": [
         "--from",
-        "deploybot-merge-queue[mcp] @ git+https://github.com/Forward-Future/DeployBot.git@de0819770c2496b5048488c2a6a207be0378af06",
+        "deploybot-merge-queue[mcp] @ git+https://github.com/Forward-Future/DeployBot.git@73004ea7c9dcb81e7f1281c0687aea0897d1571d",
         "deploybot-mcp"
       ]
     }

adapters/codex/agent-merge-queue/.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "deploybot",
-  "version": "0.2.23",
+  "version": "0.2.24",
   "description": "Coordinate exact-head pull requests through verified deployment and thread notification",
   "author": {
     "name": "DeployBot contributors"

adapters/cursor/.cursor/mcp.json

@@ -4,7 +4,7 @@
       "command": "uvx",
       "args": [
         "--from",
-        "deploybot-merge-queue[mcp] @ git+https://github.com/Forward-Future/DeployBot.git@de0819770c2496b5048488c2a6a207be0378af06",
+        "deploybot-merge-queue[mcp] @ git+https://github.com/Forward-Future/DeployBot.git@73004ea7c9dcb81e7f1281c0687aea0897d1571d",
         "deploybot-mcp"
       ]
     }

docs/reference.md

@@ -1,7 +1,7 @@
 # DeployBot reference
 
 This reference describes the CLI, MCP server, policy file, and GitHub Action in
-DeployBot v0.2.23. GitHub labels and authenticated comments are the durable state;
+DeployBot v0.2.24. GitHub labels and authenticated comments are the durable state;
 the CLI and MCP tools are two interfaces to the same operations.
 
 ## CLI

examples/github-workflow.yml

@@ -77,8 +77,8 @@ jobs:
         with:
           ref: ${{ github.event.repository.default_branch }}
           persist-credentials: false
-      # v0.2.23 implementation; keep the full commit for privileged workflows.
-      - uses: Forward-Future/DeployBot@de0819770c2496b5048488c2a6a207be0378af06
+      # v0.2.24 implementation; keep the full commit for privileged workflows.
+      - uses: Forward-Future/DeployBot@73004ea7c9dcb81e7f1281c0687aea0897d1571d
         with:
           # PR and review events reconcile quickly. Only release-owner events
           # stay attached to cumulative main through CI and deployment.

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "deploybot-merge-queue"
-version = "0.2.23"
+version = "0.2.24"
 description = "DeployBot: a provider-neutral GitHub merge queue for coding agents"
 readme = "README.md"
 license = "MIT"

src/agent_merge_queue/__init__.py

@@ -1,3 +1,3 @@
 """DeployBot: a provider-neutral GitHub merge queue for coding agents."""
 
-__version__ = "0.2.23"
+__version__ = "0.2.24"

src/agent_merge_queue/cli.py

@@ -1199,6 +1199,16 @@ def pull_head(self, number: int) -> dict[str, str]:
             "state": str(value.get("state") or ""),
         }
 
+    def reopen_pull_request(self, number: int) -> None:
+        self._json(
+            "api",
+            "--method",
+            "PATCH",
+            f"repos/{self.repository}/pulls/{number}",
+            "-f",
+            "state=open",
+        )
+
     def source_deploy_authorized(self, number: int, expected_head: str) -> bool:
         source = self._json(
             "pr",
@@ -3512,6 +3522,12 @@ def record_integration_conflict_repair(
         source_heads=frozen_heads,
         source_pull_requests=frozen_numbers,
     )
+    for source_number in frozen_numbers:
+        if source_number == owner.number:
+            continue
+        source_labels = client.labels(source_number)
+        if client.config.blocked_label not in source_labels:
+            client.add_label(source_number, client.config.blocked_label)
     result["repair_owner"] = {
         "pull_request": owner.number,
         "provider": repair.get("provider"),
@@ -3587,6 +3603,12 @@ def record_integration_ci_repair(
         source_heads=frozen_heads,
         source_pull_requests=frozen_numbers,
     )
+    for source_number in frozen_numbers:
+        if source_number == owner.number:
+            continue
+        source_labels = client.labels(source_number)
+        if client.config.blocked_label not in source_labels:
+            client.add_label(source_number, client.config.blocked_label)
     # Publish the terminal integration marker only after the durable repair
     # handoff exists. If source reads or notification writes fail, the clean
     # marker remains retryable on the next reconciliation.
@@ -3750,14 +3772,19 @@ def evaluate(
             )
             return "blocked", {"number": number, "reason": reason}, entry
         if client.config.blocked_label in entry.labels:
-            # A conflict repair targets the cumulative PR while it is open.
-            # If that integration disappears, active_integration_sources() no
-            # longer suppresses this source, so release the integration-owned
-            # hold and let the original exact-head intent recover normally.
-            integration_repair_abandoned = bool(
-                repair and repair.get("repair_pull_request")
-            )
-            if repair_marker_is_transitional(repair) or integration_repair_abandoned:
+            if repair and repair.get("repair_pull_request"):
+                return (
+                    "waiting",
+                    {
+                        "number": number,
+                        "reasons": [
+                            "cumulative integration repair must resume before any "
+                            "frozen source can merge"
+                        ],
+                    },
+                    entry,
+                )
+            if repair_marker_is_transitional(repair):
                 if deployment_repair_required(entry):
                     reason = "; ".join(entry.reasons or ["blocked"])
                     repair = record_repair(client, entry, intent, reason)
@@ -4199,6 +4226,9 @@ def command_resume(client: GitHub, selector: str | None) -> None:
     )
     if integration:
         pull = client.pull_head(number)
+        if pull.get("state") == "CLOSED":
+            client.reopen_pull_request(number)
+            pull = client.pull_head(number)
         branch = str(pull.get("branch") or "")
         head_sha = str(pull.get("head_sha") or "")
         known_checks = completed_integration_ci_checks(

tests/test_cli.py

@@ -3993,7 +3993,7 @@ def test_record_repair_replaces_packet_missing_frozen_members(
         self.assertEqual(refreshed["source_heads"]["8"], "8" * 40)
         client.comment.assert_called_once_with(7, repair_body(refreshed))
 
-    def test_promote_recovers_owner_after_conflicted_integration_closes(self) -> None:
+    def test_promote_keeps_owner_blocked_if_integration_closes(self) -> None:
         ready = entry(1)
         ready.labels = ["deploy-requested", "merge-queue-blocked"]
         intent_comment = {
@@ -4036,9 +4036,11 @@ def test_promote_recovers_owner_after_conflicted_integration_closes(self) -> Non
         with redirect_stdout(io.StringIO()):
             result = command_promote(client)
 
-        self.assertEqual(result["promoted"], [1])
-        client.remove_label.assert_any_call(1, "merge-queue-blocked")
-        client.add_label.assert_called_with(1, "merge-queue")
+        self.assertEqual(result["promoted"], [])
+        self.assertEqual(result["waiting"][0]["number"], 1)
+        self.assertIn("cumulative integration repair", result["waiting"][0]["reasons"][0])
+        client.remove_label.assert_not_called()
+        client.add_label.assert_not_called()
 
     def test_promote_does_not_hold_resumed_repair_against_itself(self) -> None:
         ready = entry(1)
@@ -4799,7 +4801,10 @@ def test_conflicted_integration_elects_one_tracked_repair_owner(self) -> None:
             repair["source_heads"],
             {"1": first.head_sha, "2": second.head_sha},
         )
-        client.add_label.assert_called_once_with(1, CONFIG.blocked_label)
+        self.assertEqual(
+            [value.args for value in client.add_label.call_args_list],
+            [(1, CONFIG.blocked_label), (2, CONFIG.blocked_label)],
+        )
         self.assertTrue(
             any(
                 "deploybot resume 99" in value.args[1]
@@ -4942,6 +4947,52 @@ def test_failed_integration_stays_retryable_until_repair_owner_exists(
         client.comment.assert_not_called()
         client.add_label.assert_not_called()
 
+    def test_failed_integration_blocks_every_frozen_source(self) -> None:
+        owner = entry(1, "shared.py")
+        peer = entry(2, "shared.py")
+        intent = {
+            "id": 1,
+            "created_at": "2026-06-20T00:00:00Z",
+            "user": {"login": "trusted"},
+            "body": intent_body(
+                intent_id="intent-1",
+                state="requested",
+                requested_at="2026-06-20T00:00:00Z",
+                requested_head=owner.head_sha,
+                provider="codex",
+                thread_id="thread-1",
+            ),
+        }
+        result = {
+            "pull_request": 99,
+            "integration": {
+                "batch_id": "batch-1",
+                "conflict": None,
+                "heads": {"1": owner.head_sha, "2": peer.head_sha},
+                "pull_requests": [1, 2],
+            },
+            "reason": "integration PR #99 CI failed: CI",
+            "state": "blocked",
+        }
+        client = Mock()
+        client.config = CONFIG
+        client.trusted_logins = {"trusted"}
+        client.snapshot.side_effect = [owner]
+        client.comments.return_value = [intent]
+        client.labels.return_value = set()
+        client.base_sha.return_value = "b" * 40
+
+        record_integration_ci_repair(client, result)
+
+        self.assertEqual(
+            [value.args for value in client.add_label.call_args_list],
+            [
+                (1, CONFIG.blocked_label),
+                (2, CONFIG.blocked_label),
+                (99, CONFIG.blocked_label),
+            ],
+        )
+
     def test_resumed_integration_finishes_source_label_delegation(self) -> None:
         integration_head = "9" * 40
         marker = {
@@ -4965,11 +5016,18 @@ def test_resumed_integration_finishes_source_label_delegation(self) -> None:
         client.coordinator_logins = {"coordinator"}
         client.resolve_pr.return_value = 99
         client.comments.return_value = comments
-        client.pull_head.return_value = {
-            "branch": "deploybot/integration/batch-1",
-            "head_sha": integration_head,
-            "state": "OPEN",
-        }
+        client.pull_head.side_effect = [
+            {
+                "branch": "deploybot/integration/batch-1",
+                "head_sha": integration_head,
+                "state": "CLOSED",
+            },
+            {
+                "branch": "deploybot/integration/batch-1",
+                "head_sha": integration_head,
+                "state": "OPEN",
+            },
+        ]
         client.workflow_runs_for_branch.return_value = [
             {
                 "id": 7,
@@ -5000,6 +5058,7 @@ def test_resumed_integration_finishes_source_label_delegation(self) -> None:
         with redirect_stdout(io.StringIO()):
             command_resume(client, "99")
 
+        client.reopen_pull_request.assert_called_once_with(99)
         client.add_label.assert_called_once_with(99, CONFIG.queue_label)
         self.assertEqual(
             [value.args for value in client.remove_label.call_args_list],

tests/test_skill.py

@@ -8,7 +8,7 @@
 
 ROOT = Path(__file__).resolve().parents[1]
 CANONICAL = ROOT / "skills" / "deploybot" / "SKILL.md"
-RELEASE_COMMIT = "de0819770c2496b5048488c2a6a207be0378af06"
+RELEASE_COMMIT = "73004ea7c9dcb81e7f1281c0687aea0897d1571d"
 CHECKOUT_COMMIT = "9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0"