diff --git a/docs/RELEASE_CONSOLE.md b/docs/RELEASE_CONSOLE.md
index 6c09bd7..ca8c3a1 100644
--- a/docs/RELEASE_CONSOLE.md
+++ b/docs/RELEASE_CONSOLE.md
@@ -63,6 +63,16 @@ these digests detect accidental or manual corruption, not an attacker who can
 replace the private record and recompute its checksums. Changing a target,
 channel, or gate input requires a new run.
 
+Creation requires a caller-generated `request_id`. Its SHA-256 fingerprint is
+the private, workspace-scoped durable mapping to exactly one run; the raw
+identifier is never persisted. Repeating the same identifier with the same
+immutable inputs returns that run without rebuilding the plan, even if the
+live dashboard has since drifted. Reusing it with different inputs fails
+closed. The browser keeps an uncertain create identifier in session storage,
+replays it after reload, and selects the known run returned by the server. A
+successful create remains shown as saved if only the subsequent list refresh
+fails.
+
 Run records survive console restarts, but remain local operator state rather
 than a signed release artifact or the system audit log. The default location is
 `$XDG_STATE_HOME/govoplan/release-console/workspace-<sha256>/release-runs/`, or
@@ -79,8 +89,11 @@ rewritten nor automatically quarantined.
 The full resolved-workspace SHA-256 is part of both the private storage
 namespace and immutable input snapshot. Every list, read, and state transition
 checks it. An alternate workspace therefore cannot list or resume another
-workspace's runs even when an embedding/test `run_state_root` override is
-shared accidentally.
+workspace's runs. This remains true for an embedding/test `run_state_root`
+override: the server always appends
+`workspace-<full-sha256>/release-runs/` rather than treating the override as a
+shared record directory. A corrupt record from one workspace therefore cannot
+leak even its identifier or an integrity error into another workspace.
 
 Each frozen plan step has an explicit `pending`, `running`, `succeeded`,
 `failed`, or `interrupted` state. Plan order remains a prerequisite: a later
@@ -94,6 +107,15 @@ enum event type, timestamp, step identifier, and bounded result code—never
 commands, process output, confirmation text, credentials, bearer tokens, or
 signing material.
 
+Before a step can enter `running`, the store reserves the two command-ledger
+slots needed for worst-case recovery. An explicit resume consumes one slot and
+is accepted only while a persisted attempt is actually running. A mutating
+attempt always retains its final `effect_absent` or `effect_succeeded` slot;
+`unresolved` may be recorded at most once for that attempt and only when an
+additional slot is available. Read-only interruption and known failure retain
+the slot needed to prepare a retry. Capacity exhaustion is therefore detected
+before starting an effect rather than stranding an ambiguous attempt.
+
 An explicit resume after a process restart converts every persisted `running`
 step to `interrupted`; it never guesses whether an external effect happened.
 An interrupted read-only step can be prepared for retry. An interrupted
@@ -105,10 +127,15 @@ effect, and `unresolved` keeps the run blocked. Each outcome emits a bounded,
 code-only state event. A known failed attempt can likewise be prepared for
 retry. The UI keeps unavailable controls visible and disabled.
 
+The browser likewise retains the request identifier for an uncertain
+resume/retry/reconciliation response and reuses it on the operator's retry.
+Identifiers are cleared only after the server returns the resulting run state.
+
 The run API is covered by the same local console token middleware as every
 other `/api/` route:
 
-- `POST /api/release-runs` rebuilds and freezes a selective plan.
+- `POST /api/release-runs` requires `request_id`, then idempotently rebuilds and
+  freezes a selective plan only when that creation is not already known.
 - `GET /api/release-runs` lists bounded summaries; unreadable entries are
   reported as unavailable instead of being silently omitted.
 - `GET /api/release-runs/{run_id}` reads and verifies one exact record.
diff --git a/tests/test_release_run_state.py b/tests/test_release_run_state.py
index f0c5cea..3cf646e 100644
--- a/tests/test_release_run_state.py
+++ b/tests/test_release_run_state.py
@@ -1,9 +1,13 @@
 from __future__ import annotations
 
 from concurrent.futures import ThreadPoolExecutor
+from itertools import count
 import json
 import os
 from pathlib import Path
+import re
+import shutil
+import subprocess
 import sys
 import tempfile
 import threading
@@ -11,6 +15,7 @@ import unittest
 from unittest.mock import patch
 
 from fastapi.testclient import TestClient
+from filelock import FileLock
 
 
 META_ROOT = Path(__file__).resolve().parents[1]
@@ -34,6 +39,7 @@ from server.app import (  # noqa: E402
 
 
 WORKSPACE_FINGERPRINT = "a" * 64
+CREATE_REQUESTS = count(1)
 
 
 class ReleaseRunStoreTests(unittest.TestCase):
@@ -49,7 +55,8 @@ class ReleaseRunStoreTests(unittest.TestCase):
         self.temporary.cleanup()
 
     def test_create_persists_private_immutable_plan_and_ordered_steps(self) -> None:
-        run = self.store.create(
+        run = create_run(
+            self.store,
             input_snapshot=run_input(), plan_snapshot=release_plan()
         )
 
@@ -60,7 +67,7 @@ class ReleaseRunStoreTests(unittest.TestCase):
         record_path = self.root / f"{run['run_id']}.json"
 
         self.assertEqual("govoplan.release-run", reopened["schema"])
-        self.assertEqual(2, reopened["schema_version"])
+        self.assertEqual(3, reopened["schema_version"])
         self.assertEqual(
             {"govoplan-core": "1.2.3", "govoplan-files": "1.2.4"},
             reopened["immutable"]["input"]["repo_versions"],
@@ -78,10 +85,43 @@ class ReleaseRunStoreTests(unittest.TestCase):
         plan = release_plan()
         plan["dry_run_steps"] = []
         with self.assertRaisesRegex(ReleaseRunConflict, "step collection"):
-            self.store.create(input_snapshot=run_input(), plan_snapshot=plan)
+            create_run(self.store, input_snapshot=run_input(), plan_snapshot=plan)
+
+    def test_create_request_is_durable_idempotent_and_input_bound(self) -> None:
+        request_id = "durable-create-request-0001"
+        created = create_run(self.store, request_id=request_id)
+        changed_plan = release_plan()
+        changed_plan["dry_run_steps"][0]["title"] = "Dashboard drift"  # type: ignore[index]
+
+        replayed = create_run(
+            self.store,
+            request_id=request_id,
+            plan_snapshot=changed_plan,
+        )
+        found = self.store.find_created_run(
+            request_id=request_id,
+            input_snapshot=run_input(),
+        )
+        self.assertEqual(created["run_id"], replayed["run_id"])
+        self.assertEqual(created["run_id"], found["run_id"] if found else None)
+        self.assertEqual(
+            "Inspect Core",
+            replayed["immutable"]["plan"]["dry_run_steps"][0]["title"],
+        )
+        self.assertEqual(1, len(list(self.root.glob("*.json"))))
+
+        changed_input = run_input()
+        changed_input["public_catalog"] = True
+        with self.assertRaisesRegex(ReleaseRunConflict, "different inputs"):
+            create_run(
+                self.store,
+                request_id=request_id,
+                input_snapshot=changed_input,
+            )
 
     def test_attempt_transitions_are_idempotent_and_strictly_ordered(self) -> None:
-        run_id = self.store.create(
+        run_id = create_run(
+            self.store,
             input_snapshot=run_input(), plan_snapshot=release_plan()
         )["run_id"]
 
@@ -122,7 +162,8 @@ class ReleaseRunStoreTests(unittest.TestCase):
         self.assertTrue(finished["state"]["steps"][1]["available"])
 
     def test_failed_attempt_and_delayed_start_replays_are_idempotent(self) -> None:
-        run_id = self.store.create(
+        run_id = create_run(
+            self.store,
             input_snapshot=run_input(), plan_snapshot=release_plan()
         )["run_id"]
         attempt_id = "failed-attempt-replay-0001"
@@ -155,77 +196,129 @@ class ReleaseRunStoreTests(unittest.TestCase):
             len(retried["state"]["events"]), len(delayed["state"]["events"])
         )
 
-    def test_delayed_resume_request_is_idempotent_for_the_run_lifetime(self) -> None:
-        run_id = self.store.create(
+    def test_resume_requires_a_running_attempt_and_only_records_an_effect(self) -> None:
+        run_id = create_run(
+            self.store,
             input_snapshot=run_input(), plan_snapshot=release_plan()
         )["run_id"]
         original_request = "resume-original-request-0001"
-        self.store.resume(run_id, request_id=original_request)
-        for index in range(130):
-            self.store.resume(
-                run_id, request_id=f"resume-intervening-request-{index:04d}"
-            )
-        running = self.store.start_step(
+        with self.assertRaisesRegex(ReleaseRunConflict, "only be resumed"):
+            self.store.resume(run_id, request_id=original_request)
+        before_start = self.store.get(run_id)
+
+        self.store.start_step(
             run_id,
             "core:preflight",
             attempt_id="resume-delayed-attempt-0001",
         )
+        resumed = self.store.resume(run_id, request_id=original_request)
         duplicate = self.store.resume(run_id, request_id=original_request)
-        self.assertEqual("running", duplicate["state"]["steps"][0]["state"])
+        self.assertEqual("interrupted", duplicate["state"]["steps"][0]["state"])
         self.assertEqual(
-            len(running["state"]["events"]), len(duplicate["state"]["events"])
+            len(resumed["state"]["events"]), len(duplicate["state"]["events"])
         )
+        raw = json.loads(
+            (self.root / f"{run_id}.json").read_text(encoding="utf-8")
+        )
+        self.assertEqual(1, len(raw["state"]["processed_requests"]))
+        self.assertLess(len(before_start["state"]["events"]), len(resumed["state"]["events"]))
 
-    def test_command_and_attempt_capacity_fail_closed_without_eviction(self) -> None:
-        run_id = self.store.create(
-            input_snapshot=run_input(), plan_snapshot=release_plan()
-        )["run_id"]
-        with patch("govoplan_release.release_run.MAX_COMMANDS", 2):
-            self.store.resume(run_id, request_id="capacity-resume-request-0001")
-            self.store.resume(run_id, request_id="capacity-resume-request-0002")
-            with self.assertRaisesRegex(ReleaseRunConflict, "safety limit"):
-                self.store.resume(run_id, request_id="capacity-resume-request-0003")
-            duplicate = self.store.resume(
-                run_id, request_id="capacity-resume-request-0001"
-            )
-        self.assertEqual("planned", duplicate["state"]["status"])
-
-        attempt_run_id = self.store.create(
+    def test_attempt_start_reserves_capacity_for_ambiguous_effect_recovery(self) -> None:
+        saturated_run_id = create_run(
+            self.store,
             input_snapshot=run_input(), plan_snapshot=release_plan()
         )["run_id"]
         with patch("govoplan_release.release_run.MAX_COMMANDS", 1):
-            self.store.start_step(
-                attempt_run_id,
-                "core:preflight",
-                attempt_id="capacity-attempt-0001",
-            )
-            self.store.finish_step(
-                attempt_run_id,
-                "core:preflight",
-                attempt_id="capacity-attempt-0001",
-                succeeded=False,
-                result_code="preflight_failed",
-            )
-            self.store.retry_step(
-                attempt_run_id,
-                "core:preflight",
-                request_id="capacity-retry-request-0001",
-            )
-            with self.assertRaisesRegex(ReleaseRunConflict, "attempt history"):
+            with self.assertRaisesRegex(ReleaseRunConflict, "recovery capacity"):
                 self.store.start_step(
-                    attempt_run_id,
+                    saturated_run_id,
                     "core:preflight",
-                    attempt_id="capacity-attempt-0002",
+                    attempt_id="capacity-prestart-attempt-0001",
                 )
-            delayed_attempt = self.store.start_step(
-                attempt_run_id,
-                "core:preflight",
-                attempt_id="capacity-attempt-0001",
+        untouched = self.store.get(saturated_run_id)
+        self.assertEqual("pending", untouched["state"]["steps"][0]["state"])
+        self.assertEqual(0, untouched["state"]["steps"][0]["attempt_count"])
+
+        terminal_run_id = create_run(
+            self.store,
+            input_snapshot=run_input(),
+            plan_snapshot=mutating_first_plan(),
+        )["run_id"]
+        with patch("govoplan_release.release_run.MAX_COMMANDS", 2):
+            self.store.start_step(
+                terminal_run_id, "core:tag", attempt_id="capacity-terminal-attempt"
             )
-        self.assertEqual("pending", delayed_attempt["state"]["steps"][0]["state"])
+            self.store.resume(
+                terminal_run_id, request_id="capacity-terminal-resume-0001"
+            )
+            with self.assertRaisesRegex(ReleaseRunConflict, "safety limit"):
+                self.store.reconcile_step(
+                    terminal_run_id,
+                    "core:tag",
+                    request_id="capacity-unresolved-request-0001",
+                    outcome="unresolved",
+                    confirmation="RECONCILE",
+                )
+            completed = self.store.reconcile_step(
+                terminal_run_id,
+                "core:tag",
+                request_id="capacity-terminal-request-0001",
+                outcome="effect_succeeded",
+                confirmation="RECONCILE",
+            )
+        self.assertEqual("completed", completed["state"]["status"])
+
+    def test_unresolved_reconciliation_is_bounded_and_preserves_terminal_slot(self) -> None:
+        run_id = create_run(
+            self.store,
+            input_snapshot=run_input(),
+            plan_snapshot=mutating_first_plan(),
+        )["run_id"]
+        with patch("govoplan_release.release_run.MAX_COMMANDS", 3):
+            self.store.start_step(
+                run_id, "core:tag", attempt_id="capacity-unresolved-attempt"
+            )
+            self.store.resume(
+                run_id, request_id="capacity-unresolved-resume-0001"
+            )
+            unresolved = self.store.reconcile_step(
+                run_id,
+                "core:tag",
+                request_id="capacity-unresolved-request-0001",
+                outcome="unresolved",
+                confirmation="RECONCILE",
+            )
+            duplicate = self.store.reconcile_step(
+                run_id,
+                "core:tag",
+                request_id="capacity-unresolved-request-0001",
+                outcome="unresolved",
+                confirmation="RECONCILE",
+            )
+            self.assertEqual(
+                len(unresolved["state"]["events"]),
+                len(duplicate["state"]["events"]),
+            )
+            with self.assertRaisesRegex(ReleaseRunConflict, "already recorded"):
+                self.store.reconcile_step(
+                    run_id,
+                    "core:tag",
+                    request_id="capacity-unresolved-request-0002",
+                    outcome="unresolved",
+                    confirmation="RECONCILE",
+                )
+            completed = self.store.reconcile_step(
+                run_id,
+                "core:tag",
+                request_id="capacity-final-request-0001",
+                outcome="effect_succeeded",
+                confirmation="RECONCILE",
+            )
+        self.assertEqual("completed", completed["state"]["status"])
 
     def test_reopen_and_resume_preserve_ambiguous_mutating_effect(self) -> None:
-        run_id = self.store.create(
+        run_id = create_run(
+            self.store,
             input_snapshot=run_input(), plan_snapshot=mutating_first_plan()
         )["run_id"]
         self.store.start_step(run_id, "core:tag", attempt_id="attempt-tag-0001")
@@ -281,7 +374,8 @@ class ReleaseRunStoreTests(unittest.TestCase):
         )
 
     def test_reconciliation_can_record_unresolved_or_verified_success(self) -> None:
-        run_id = self.store.create(
+        run_id = create_run(
+            self.store,
             input_snapshot=run_input(), plan_snapshot=mutating_first_plan()
         )["run_id"]
         self.store.start_step(run_id, "core:tag", attempt_id="reconcile-attempt-0001")
@@ -322,7 +416,8 @@ class ReleaseRunStoreTests(unittest.TestCase):
         self.assertEqual("completed", succeeded["state"]["status"])
 
     def test_tampered_record_fails_closed_and_is_not_rewritten(self) -> None:
-        run_id = self.store.create(
+        run_id = create_run(
+            self.store,
             input_snapshot=run_input(), plan_snapshot=release_plan()
         )["run_id"]
         path = self.root / f"{run_id}.json"
@@ -338,7 +433,8 @@ class ReleaseRunStoreTests(unittest.TestCase):
         self.assertEqual("unavailable", self.store.list()[0]["status"])
 
     def test_broad_record_mode_and_symlinked_root_fail_closed(self) -> None:
-        run_id = self.store.create(
+        run_id = create_run(
+            self.store,
             input_snapshot=run_input(), plan_snapshot=release_plan()
         )["run_id"]
         path = self.root / f"{run_id}.json"
@@ -357,8 +453,58 @@ class ReleaseRunStoreTests(unittest.TestCase):
                 expected_workspace_fingerprint=WORKSPACE_FINGERPRINT,
             ).list()
 
+    def test_symlinked_lock_and_lock_io_errors_fail_closed(self) -> None:
+        self.store.list()
+        lock_path = self.root / ".release-runs.lock"
+        lock_path.unlink()
+        target = self.root / "lock-target"
+        target.write_text("do not follow", encoding="utf-8")
+        lock_path.symlink_to(target)
+        with self.assertRaisesRegex(ReleaseRunCorrupt, "lock path"):
+            self.store.list()
+
+        lock_path.unlink()
+        with patch(
+            "govoplan_release.release_run.FileLock.acquire",
+            side_effect=PermissionError("denied"),
+        ):
+            with self.assertRaisesRegex(ReleaseRunCorrupt, "acquired safely"):
+                self.store.list()
+
+        original_release = FileLock.release
+
+        def fail_operator_release(
+            lock: FileLock, *args: object, **kwargs: object
+        ) -> None:
+            if kwargs.get("force") is True:
+                original_release(lock, *args, **kwargs)  # type: ignore[arg-type]
+                return
+            raise PermissionError("denied")
+
+        with patch.object(
+            FileLock,
+            "release",
+            autospec=True,
+            side_effect=fail_operator_release,
+        ):
+            with self.assertRaisesRegex(ReleaseRunCorrupt, "released safely"):
+                self.store.list()
+
+    def test_atomic_write_oserror_is_normalized_and_leaves_no_record(self) -> None:
+        with patch(
+            "govoplan_release.release_run.os.replace",
+            side_effect=PermissionError("denied"),
+        ):
+            with self.assertRaisesRegex(ReleaseRunCorrupt, "persisted safely"):
+                create_run(
+                    self.store,
+                    request_id="atomic-write-create-request-0001",
+                )
+        self.assertEqual([], list(self.root.glob("*.json")))
+
     def test_state_specific_fields_and_boolean_counts_fail_closed(self) -> None:
-        run_id = self.store.create(
+        run_id = create_run(
+            self.store,
             input_snapshot=run_input(), plan_snapshot=release_plan()
         )["run_id"]
         path = self.root / f"{run_id}.json"
@@ -372,7 +518,8 @@ class ReleaseRunStoreTests(unittest.TestCase):
             self.store.get(run_id)
 
     def test_valid_looking_mutable_state_tampering_fails_record_checksum(self) -> None:
-        run_id = self.store.create(
+        run_id = create_run(
+            self.store,
             input_snapshot=run_input(), plan_snapshot=mutating_first_plan()
         )["run_id"]
         self.store.start_step(run_id, "core:tag", attempt_id="checksum-attempt-0001")
@@ -396,7 +543,8 @@ class ReleaseRunStoreTests(unittest.TestCase):
         self.assertEqual(tampered, path.read_bytes())
 
     def test_impossible_step_chronology_fails_even_with_recomputed_checksum(self) -> None:
-        run_id = self.store.create(
+        run_id = create_run(
+            self.store,
             input_snapshot=run_input(), plan_snapshot=release_plan()
         )["run_id"]
         self.store.start_step(
@@ -421,9 +569,13 @@ class ReleaseRunStoreTests(unittest.TestCase):
             self.store.get(run_id)
 
     def test_record_and_retained_event_timestamps_are_strictly_ordered(self) -> None:
-        run_id = self.store.create(
+        run_id = create_run(
+            self.store,
             input_snapshot=run_input(), plan_snapshot=release_plan()
         )["run_id"]
+        self.store.start_step(
+            run_id, "core:preflight", attempt_id="timestamp-attempt-0001"
+        )
         self.store.resume(run_id, request_id="timestamp-resume-request-0001")
         path = self.root / f"{run_id}.json"
         payload = json.loads(path.read_text(encoding="utf-8"))
@@ -439,7 +591,8 @@ class ReleaseRunStoreTests(unittest.TestCase):
             self.store.get(run_id)
 
     def test_attempt_count_must_match_lifetime_attempt_history(self) -> None:
-        run_id = self.store.create(
+        run_id = create_run(
+            self.store,
             input_snapshot=run_input(), plan_snapshot=release_plan()
         )["run_id"]
         self.store.start_step(
@@ -456,7 +609,8 @@ class ReleaseRunStoreTests(unittest.TestCase):
             self.store.get(run_id)
 
     def test_workspace_fingerprint_is_enforced_for_every_store_operation(self) -> None:
-        run_id = self.store.create(
+        run_id = create_run(
+            self.store,
             input_snapshot=run_input(), plan_snapshot=release_plan()
         )["run_id"]
         foreign = ReleaseRunStore(
@@ -470,7 +624,11 @@ class ReleaseRunStoreTests(unittest.TestCase):
         with self.assertRaises(ReleaseRunNotFound):
             foreign.resume(run_id, request_id="foreign-resume-request-0001")
         with self.assertRaisesRegex(ReleaseRunConflict, "different workspace"):
-            foreign.create(input_snapshot=run_input(), plan_snapshot=release_plan())
+            create_run(
+                foreign,
+                input_snapshot=run_input(),
+                plan_snapshot=release_plan(),
+            )
 
     def test_list_limit_is_applied_after_foreign_workspace_records_are_skipped(
         self,
@@ -479,7 +637,8 @@ class ReleaseRunStoreTests(unittest.TestCase):
             "govoplan_release.release_run._timestamp",
             return_value="2026-07-22T00:00:00Z",
         ):
-            local_run = self.store.create(
+            local_run = create_run(
+                self.store,
                 input_snapshot=run_input(), plan_snapshot=release_plan()
             )
         foreign_store = ReleaseRunStore(
@@ -492,7 +651,8 @@ class ReleaseRunStoreTests(unittest.TestCase):
             "govoplan_release.release_run._timestamp",
             return_value="2026-07-22T00:00:01Z",
         ):
-            foreign_store.create(
+            create_run(
+                foreign_store,
                 input_snapshot=foreign_input, plan_snapshot=release_plan()
             )
 
@@ -500,7 +660,8 @@ class ReleaseRunStoreTests(unittest.TestCase):
         self.assertEqual([local_run["run_id"]], [item["run_id"] for item in listed])
 
     def test_concurrent_claims_have_one_winner_and_no_lost_update(self) -> None:
-        run_id = self.store.create(
+        run_id = create_run(
+            self.store,
             input_snapshot=run_input(), plan_snapshot=release_plan()
         )["run_id"]
         barrier = threading.Barrier(2)
@@ -535,7 +696,8 @@ class ReleaseRunStoreTests(unittest.TestCase):
         )
 
     def test_event_history_is_bounded_and_command_history_is_not_evicted(self) -> None:
-        run_id = self.store.create(
+        run_id = create_run(
+            self.store,
             input_snapshot=run_input(), plan_snapshot=release_plan()
         )["run_id"]
         for index in range(90):
@@ -588,7 +750,7 @@ class ReleaseRunApiTests(unittest.TestCase):
                 created_response = client.post(
                     "/api/release-runs",
                     headers=headers,
-                    json=run_input(),
+                    json=api_run_input(request_id="api-create-request-0001"),
                 )
                 self.assertEqual(200, created_response.status_code)
                 created = created_response.json()
@@ -623,6 +785,59 @@ class ReleaseRunApiTests(unittest.TestCase):
             self.assertEqual(200, retried.status_code)
             self.assertEqual("pending", retried.json()["state"]["steps"][0]["state"])
 
+    def test_create_requires_request_id_and_replays_before_plan_rebuild(self) -> None:
+        with tempfile.TemporaryDirectory() as temp_dir:
+            app = create_app(
+                workspace_root=Path(temp_dir),
+                run_state_root=Path(temp_dir) / "runs",
+                token="token",
+            )
+            headers = {"X-Release-Console-Token": "token"}
+            dashboard = patch("server.app.build_dashboard", return_value=object())
+            planner = patch(
+                "server.app.build_selective_release_plan",
+                return_value=release_plan(),
+            )
+            with dashboard as build_dashboard, planner as build_plan, TestClient(
+                app
+            ) as client:
+                missing = client.post(
+                    "/api/release-runs",
+                    headers=headers,
+                    json={
+                        key: value
+                        for key, value in api_run_input().items()
+                        if key != "request_id"
+                    },
+                )
+                payload = api_run_input(request_id="api-idempotent-create-0001")
+                created = client.post(
+                    "/api/release-runs", headers=headers, json=payload
+                )
+                self.assertEqual(200, created.status_code)
+                build_dashboard.reset_mock()
+                build_plan.reset_mock()
+
+                replayed = client.post(
+                    "/api/release-runs", headers=headers, json=payload
+                )
+                mismatched = client.post(
+                    "/api/release-runs",
+                    headers=headers,
+                    json=payload | {"public_catalog": True},
+                )
+
+            self.assertEqual(422, missing.status_code)
+            self.assertEqual(200, replayed.status_code)
+            self.assertEqual(created.json()["run_id"], replayed.json()["run_id"])
+            self.assertEqual(409, mismatched.status_code)
+            self.assertIn("different inputs", mismatched.json()["detail"])
+            build_dashboard.assert_not_called()
+            build_plan.assert_not_called()
+            self.assertEqual(
+                1, len(list(app.state.release_runs.root.glob("*.json")))
+            )
+
     def test_create_rejects_unknown_or_unresolved_repository(self) -> None:
         with tempfile.TemporaryDirectory() as temp_dir:
             root = Path(temp_dir) / "runs"
@@ -641,6 +856,7 @@ class ReleaseRunApiTests(unittest.TestCase):
                     "/api/release-runs",
                     headers={"X-Release-Console-Token": "token"},
                     json={
+                        "request_id": "api-unknown-create-request-0001",
                         "channel": "stable",
                         "repo_versions": {"govoplan-unknown": "1.0.0"},
                         "public_catalog": False,
@@ -654,19 +870,20 @@ class ReleaseRunApiTests(unittest.TestCase):
     def test_api_rejects_tampered_record_without_rewriting_it(self) -> None:
         with tempfile.TemporaryDirectory() as temp_dir:
             root = Path(temp_dir) / "runs"
-            store = ReleaseRunStore(
-                root,
-                expected_workspace_fingerprint=WORKSPACE_FINGERPRINT,
-            )
-            run_id = store.create(
-                input_snapshot=run_input(), plan_snapshot=release_plan()
-            )["run_id"]
-            path = root / f"{run_id}.json"
-            path.write_text("{}", encoding="utf-8")
-            malformed = path.read_bytes()
             app = create_app(
                 workspace_root=Path(temp_dir), run_state_root=root, token="token"
             )
+            run_id = create_run(
+                app.state.release_runs,
+                request_id="api-tamper-create-request-0001",
+                input_snapshot=run_input(
+                    workspace_fingerprint=app.state.workspace_fingerprint
+                ),
+                plan_snapshot=release_plan(),
+            )["run_id"]
+            path = app.state.release_runs.root / f"{run_id}.json"
+            path.write_text("{}", encoding="utf-8")
+            malformed = path.read_bytes()
             with TestClient(app) as client:
                 response = client.get(
                     f"/api/release-runs/{run_id}",
@@ -712,7 +929,9 @@ class ReleaseRunApiTests(unittest.TestCase):
                 TestClient(app) as client,
             ):
                 created = client.post(
-                    "/api/release-runs", headers=headers, json=run_input()
+                    "/api/release-runs",
+                    headers=headers,
+                    json=api_run_input(request_id="api-reconcile-create-0001"),
                 ).json()
                 run_id = created["run_id"]
                 app.state.release_runs.start_step(
@@ -807,14 +1026,19 @@ class ReleaseRunApiTests(unittest.TestCase):
                 response = client.post(
                     "/api/release-runs",
                     headers={"X-Release-Console-Token": "token"},
-                    json=run_input(),
+                    json=api_run_input(request_id="api-explicit-create-0001"),
                 )
             self.assertEqual(200, response.status_code)
             self.assertEqual(
                 release_workspace_fingerprint(first),
                 response.json()["immutable"]["input"]["workspace_fingerprint"],
             )
-            self.assertEqual(explicit_root.resolve(), app.state.release_runs.root)
+            self.assertEqual(
+                explicit_root.resolve()
+                / f"workspace-{release_workspace_fingerprint(first)}"
+                / "release-runs",
+                app.state.release_runs.root,
+            )
 
     def test_shared_explicit_root_does_not_cross_workspace_boundary(self) -> None:
         with tempfile.TemporaryDirectory() as temp_dir:
@@ -837,9 +1061,16 @@ class ReleaseRunApiTests(unittest.TestCase):
                 TestClient(first_app) as client,
             ):
                 created = client.post(
-                    "/api/release-runs", headers=headers, json=run_input()
+                    "/api/release-runs",
+                    headers=headers,
+                    json=api_run_input(request_id="api-shared-create-0001"),
                 ).json()
 
+            first_record = (
+                first_app.state.release_runs.root / f"{created['run_id']}.json"
+            )
+            first_record.write_text("{}", encoding="utf-8")
+
             second_app = create_app(
                 workspace_root=second, run_state_root=root, token="token"
             )
@@ -850,6 +1081,10 @@ class ReleaseRunApiTests(unittest.TestCase):
                 )
             self.assertEqual([], listed.json()["runs"])
             self.assertEqual(404, loaded.status_code)
+            self.assertNotEqual(
+                first_app.state.release_runs.root,
+                second_app.state.release_runs.root,
+            )
 
     def test_ui_and_runbook_state_tracking_boundary_are_explicit(self) -> None:
         webui = (RELEASE_ROOT / "webui" / "index.html").read_text(encoding="utf-8")
@@ -866,13 +1101,163 @@ class ReleaseRunApiTests(unittest.TestCase):
         self.assertIn("effect_succeeded", webui)
         self.assertIn("renderReleaseRunStoreUnavailable", webui)
         self.assertIn('const runsRequest = api("/api/release-runs")', webui)
+        self.assertIn("pendingReleaseRunPayload", webui)
+        self.assertIn("recoverPendingReleaseRun", webui)
+        self.assertIn("inFlightRunCommandId", webui)
+        self.assertIn("Run saved; list refresh unavailable", webui)
         self.assertIn("This foundation tracks state only.", runbook)
         self.assertIn("effect_absent", runbook)
         self.assertIn("$XDG_STATE_HOME", runbook)
         self.assertIn("same-directory", runbook)
+        self.assertIn("caller-generated `request_id`", runbook)
+        self.assertIn("reserves the two command-ledger", runbook)
+
+    def test_ui_reuses_uncertain_ids_and_separates_saved_run_list_failure(
+        self,
+    ) -> None:
+        node = shutil.which("node")
+        if node is None:
+            self.skipTest("Node.js is required for the Release Console UI check")
+        webui = (RELEASE_ROOT / "webui" / "index.html").read_text(encoding="utf-8")
+        function_names = (
+            "pendingReleaseRunPayload",
+            "readPendingReleaseRun",
+            "clearPendingReleaseRun",
+            "recoverPendingReleaseRun",
+            "submitReleaseRunCreate",
+            "refreshReleaseRunListAfterCreate",
+            "readInFlightRunCommands",
+            "inFlightRunCommandId",
+            "clearInFlightRunCommand",
+            "requestId",
+        )
+        functions = "\n".join(
+            javascript_function(webui, name) for name in function_names
+        )
+        script = (
+            """
+const assert = (condition, message) => { if (!condition) throw new Error(message); };
+const values = new Map();
+const sessionStorage = {
+  getItem: (key) => values.has(key) ? values.get(key) : null,
+  setItem: (key, value) => values.set(key, String(value)),
+  removeItem: (key) => values.delete(key),
+};
+let uuidSequence = 0;
+const window = { crypto: { randomUUID: () => `uuid-${++uuidSequence}` } };
+const pendingReleaseRunKey = "test-pending-create";
+const pendingRunCommandsKey = "test-pending-commands";
+const releaseState = { currentRun: null, runStoreAvailable: true };
+const elements = {
+  runStatus: { textContent: "" },
+  runOutput: {
+    innerHTML: "",
+    insertAdjacentHTML: (_position, html) => { elements.runOutput.innerHTML = html + elements.runOutput.innerHTML; },
+  },
+  releaseRun: { value: "" },
+};
+const pill = (label) => label;
+const escapeHtml = (value) => String(value);
+const restoreReleaseRunSelection = () => {};
+const renderReleaseRun = (run) => { elements.runOutput.innerHTML = `RUN:${run.run_id}`; };
+const renderReleaseRunList = () => {};
+let failCreate = true;
+let failList = true;
+const postedRequestIds = [];
+async function postJson(_path, payload) {
+  postedRequestIds.push(payload.request_id);
+  if (failCreate) throw new Error("uncertain transport failure");
+  return { run_id: "rr-known", immutable: { input: {} }, state: { steps: [] } };
+}
+async function api(_path) {
+  if (failList) throw new Error("list refresh failed");
+  return { runs: [] };
+}
+"""
+            + functions
+            + """
+(async () => {
+  const input = { repo_versions: { "govoplan-files": "1.2.4", "govoplan-core": "1.2.3" }, channel: "stable" };
+  const first = pendingReleaseRunPayload(input);
+  const same = pendingReleaseRunPayload({ ...input, repo_versions: { "govoplan-core": "1.2.3", "govoplan-files": "1.2.4" } });
+  assert(first.request_id === same.request_id, "same create input did not reuse its request ID");
+  const uncertain = await submitReleaseRunCreate(first, { automatic: true });
+  assert(uncertain === false, "uncertain create unexpectedly succeeded");
+  assert(readPendingReleaseRun().payload.request_id === first.request_id, "uncertain create ID was cleared");
+
+  failCreate = false;
+  const recovered = await recoverPendingReleaseRun();
+  assert(recovered === true, "pending create did not recover");
+  assert(postedRequestIds.length === 2 && postedRequestIds.every((value) => value === first.request_id), "create replay changed request ID");
+  assert(readPendingReleaseRun() === null, "successful replay did not clear pending create");
+  assert(elements.runOutput.innerHTML.includes("Run saved; list refresh unavailable"), "list failure did not preserve saved state");
+  assert(!elements.runOutput.innerHTML.includes("Run creation failed"), "list failure was labeled as create failure");
+  assert(releaseState.currentRun.run_id === "rr-known", "saved run was not retained");
+
+  const commandOne = inFlightRunCommandId("retry:rr-known:step", "retry");
+  const commandReplay = inFlightRunCommandId("retry:rr-known:step", "retry");
+  assert(commandOne === commandReplay, "uncertain command did not reuse its request ID");
+  clearInFlightRunCommand("retry:rr-known:step", commandOne);
+  const commandAfterSuccess = inFlightRunCommandId("retry:rr-known:step", "retry");
+  assert(commandAfterSuccess !== commandOne, "successful command did not clear its request ID");
+})().catch((error) => { console.error(error.stack || error.message); process.exit(1); });
+"""
+        )
+        result = subprocess.run(  # noqa: S603
+            [node, "-e", script],
+            check=False,
+            capture_output=True,
+            text=True,
+        )
+        self.assertEqual(0, result.returncode, result.stderr)
 
 
-def run_input() -> dict[str, object]:
+def create_run(
+    store: ReleaseRunStore,
+    *,
+    request_id: str | None = None,
+    input_snapshot: dict[str, object] | None = None,
+    plan_snapshot: dict[str, object] | None = None,
+) -> dict[str, object]:
+    return store.create(
+        request_id=request_id
+        or f"test-create-request-{next(CREATE_REQUESTS):06d}",
+        input_snapshot=input_snapshot if input_snapshot is not None else run_input(),
+        plan_snapshot=plan_snapshot if plan_snapshot is not None else release_plan(),
+    )
+
+
+def api_run_input(
+    *, request_id: str | None = None, **overrides: object
+) -> dict[str, object]:
+    payload = run_input()
+    payload.pop("workspace_fingerprint")
+    payload["request_id"] = request_id or (
+        f"api-create-request-{next(CREATE_REQUESTS):06d}"
+    )
+    payload.update(overrides)
+    return payload
+
+
+def javascript_function(source: str, name: str) -> str:
+    match = re.search(
+        rf"\n\s+(?:async\s+)?function\s+{re.escape(name)}\s*\(", source
+    )
+    if match is None:
+        raise AssertionError(f"JavaScript function {name} was not found")
+    following = re.search(
+        r"\n\s+(?:async\s+)?function\s+[A-Za-z_$]",
+        source[match.end() :],
+    )
+    if following is None:
+        raise AssertionError(f"JavaScript function after {name} was not found")
+    end = match.end() + following.start()
+    return source[match.start() : end].strip()
+
+
+def run_input(
+    *, workspace_fingerprint: str = WORKSPACE_FINGERPRINT
+) -> dict[str, object]:
     return {
         "channel": "stable",
         "repo_versions": {
@@ -883,7 +1268,7 @@ def run_input() -> dict[str, object]:
         "remote_tags": False,
         "public_catalog": False,
         "include_migrations": True,
-        "workspace_fingerprint": WORKSPACE_FINGERPRINT,
+        "workspace_fingerprint": workspace_fingerprint,
     }
 
 
diff --git a/tools/release/govoplan_release/release_run.py b/tools/release/govoplan_release/release_run.py
index 792aa77..f85d0b2 100644
--- a/tools/release/govoplan_release/release_run.py
+++ b/tools/release/govoplan_release/release_run.py
@@ -19,20 +19,21 @@ import stat
 import tempfile
 import threading
 from typing import Any
-import uuid
 
 from filelock import FileLock, Timeout as FileLockTimeout
 
 
 SCHEMA = "govoplan.release-run"
-SCHEMA_VERSION = 2
+SCHEMA_VERSION = 3
 MAX_RECORD_BYTES = 2 * 1024 * 1024
 MAX_PLAN_STEPS = 512
 MAX_REPOSITORY_VERSIONS = 128
 MAX_EVENTS = 256
 MAX_COMMANDS = 2_048
 MAX_LIST_LIMIT = 100
-RUN_ID_PATTERN = re.compile(r"^rr-[0-9]{8}T[0-9]{6}Z-[0-9a-f]{12}$")
+RUN_ID_PATTERN = re.compile(
+    r"^(?:rr-[0-9]{8}T[0-9]{6}Z-[0-9a-f]{12}|rr-request-[0-9a-f]{64})$"
+)
 STEP_ID_PATTERN = re.compile(r"^[A-Za-z0-9][A-Za-z0-9._:@-]{0,159}$")
 REQUEST_ID_PATTERN = re.compile(r"^[A-Za-z0-9][A-Za-z0-9._:@/-]{7,127}$")
 CHANNEL_PATTERN = re.compile(r"^[A-Za-z0-9][A-Za-z0-9._-]{0,63}$")
@@ -57,6 +58,7 @@ EVENT_TYPES = {
 ATTEMPT_OUTCOMES = {"running", "succeeded", "failed", "interrupted"}
 RECONCILIATION_OUTCOMES = {"effect_absent", "effect_succeeded", "unresolved"}
 RECONCILIATION_CONFIRMATION = "RECONCILE"
+START_RECOVERY_RESERVE = 2
 
 
 class ReleaseRunError(RuntimeError):
@@ -97,9 +99,11 @@ class ReleaseRunStore:
     def create(
         self,
         *,
+        request_id: str,
         input_snapshot: dict[str, Any],
         plan_snapshot: dict[str, Any],
     ) -> dict[str, Any]:
+        request_fingerprint = _request_fingerprint(request_id)
         immutable_input = _validated_input(input_snapshot)
         if (
             immutable_input["workspace_fingerprint"]
@@ -110,9 +114,7 @@ class ReleaseRunStore:
             )
         immutable_plan = _validated_plan(plan_snapshot)
         now = _timestamp()
-        run_id = (
-            f"rr-{now.replace('-', '').replace(':', '')[:15]}Z-{uuid.uuid4().hex[:12]}"
-        )
+        run_id = _creation_run_id(request_fingerprint)
         immutable = {
             "input": immutable_input,
             "plan": immutable_plan,
@@ -134,6 +136,7 @@ class ReleaseRunStore:
             "schema": SCHEMA,
             "schema_version": SCHEMA_VERSION,
             "run_id": run_id,
+            "create_request_fingerprint": request_fingerprint,
             "created_at": now,
             "updated_at": now,
             "immutable": immutable,
@@ -152,9 +155,42 @@ class ReleaseRunStore:
         record["record_digest"] = _record_digest(record)
         with self._locked():
             self._ensure_root()
+            existing = self._read_optional(run_id)
+            if existing is not None:
+                _validate_create_replay(
+                    existing,
+                    request_fingerprint=request_fingerprint,
+                    input_snapshot=immutable_input,
+                )
+                return _view(existing)
             self._write_new(record)
         return _view(record)
 
+    def find_created_run(
+        self, *, request_id: str, input_snapshot: dict[str, Any]
+    ) -> dict[str, Any] | None:
+        """Return an exact prior create command before rebuilding its plan."""
+
+        request_fingerprint = _request_fingerprint(request_id)
+        immutable_input = _validated_input(input_snapshot)
+        if (
+            immutable_input["workspace_fingerprint"]
+            != self.expected_workspace_fingerprint
+        ):
+            raise ReleaseRunConflict(
+                "Release run input belongs to a different workspace."
+            )
+        with self._locked():
+            record = self._read_optional(_creation_run_id(request_fingerprint))
+            if record is None:
+                return None
+            _validate_create_replay(
+                record,
+                request_fingerprint=request_fingerprint,
+                input_snapshot=immutable_input,
+            )
+            return _view(record)
+
     def get(self, run_id: str) -> dict[str, Any]:
         with self._locked():
             record = self._read(run_id)
@@ -214,10 +250,17 @@ class ReleaseRunStore:
             record = self._read(run_id)
             if _request_seen(record, request_fingerprint, "resume", None, None):
                 return _view(record)
+            running_steps = [
+                step
+                for step in record["state"]["steps"]
+                if step["state"] == "running"
+            ]
+            if not running_steps:
+                raise ReleaseRunConflict(
+                    "A release run can only be resumed while a step is running."
+                )
             _ensure_command_capacity(record)
-            for step in record["state"]["steps"]:
-                if step["state"] != "running":
-                    continue
+            for step in running_steps:
                 attempt = _attempt_for_fingerprint(
                     record, step["attempt_fingerprint"], step_id=step["id"]
                 )
@@ -232,7 +275,14 @@ class ReleaseRunStore:
                     step_id=step["id"],
                     result_code="process_interrupted",
                 )
-            _remember_request(record, request_fingerprint, "resume", None, None)
+            _remember_request(
+                record,
+                request_fingerprint,
+                "resume",
+                None,
+                None,
+                attempt_fingerprint=running_steps[0]["attempt_fingerprint"],
+            )
             _append_event(record, event_type="run_resumed")
             self._persist_update(record)
         return _view(record)
@@ -260,6 +310,7 @@ class ReleaseRunStore:
                     "Record the observed effect of an interrupted mutating step before retry."
                 )
             _ensure_command_capacity(record)
+            attempt_fingerprint = step["attempt_fingerprint"]
             retry_reason = (
                 "known_failure"
                 if step["state"] == "failed"
@@ -274,7 +325,14 @@ class ReleaseRunStore:
                     "result_code": None,
                 }
             )
-            _remember_request(record, request_fingerprint, "retry", step_id, None)
+            _remember_request(
+                record,
+                request_fingerprint,
+                "retry",
+                step_id,
+                None,
+                attempt_fingerprint=attempt_fingerprint,
+            )
             _append_event(
                 record,
                 event_type="step_retry_requested",
@@ -316,7 +374,20 @@ class ReleaseRunStore:
                 raise ReleaseRunConflict(
                     "Only an interrupted mutating step can be reconciled."
                 )
-            _ensure_command_capacity(record)
+            attempt_fingerprint = step["attempt_fingerprint"]
+            if outcome == "unresolved":
+                if any(
+                    request["operation"] == "reconcile"
+                    and request["parameter"] == "unresolved"
+                    and request["attempt_fingerprint"] == attempt_fingerprint
+                    for request in record["state"]["processed_requests"]
+                ):
+                    raise ReleaseRunConflict(
+                        "An unresolved outcome is already recorded for this attempt."
+                    )
+                _ensure_command_capacity(record, reserve_after=1)
+            else:
+                _ensure_command_capacity(record)
             if outcome == "effect_absent":
                 step.update(
                     {
@@ -336,7 +407,12 @@ class ReleaseRunStore:
                     }
                 )
             _remember_request(
-                record, request_fingerprint, "reconcile", step_id, outcome
+                record,
+                request_fingerprint,
+                "reconcile",
+                step_id,
+                outcome,
+                attempt_fingerprint=attempt_fingerprint,
             )
             _append_event(
                 record,
@@ -374,6 +450,7 @@ class ReleaseRunStore:
             available, reason = _step_available(record, step_id)
             if not available:
                 raise ReleaseRunConflict(reason)
+            _ensure_start_recovery_capacity(record)
             _ensure_attempt_capacity(record)
             step.update(
                 {
@@ -506,11 +583,22 @@ class ReleaseRunStore:
             raise ReleaseRunCorrupt("Release run record is unreadable.") from exc
         finally:
             if descriptor >= 0:
-                os.close(descriptor)
+                try:
+                    os.close(descriptor)
+                except OSError as exc:
+                    raise ReleaseRunCorrupt(
+                        "Release run record could not be closed safely."
+                    ) from exc
         _validate_record(payload, expected_run_id=run_id)
         self._assert_workspace(payload)
         return payload
 
+    def _read_optional(self, run_id: str) -> dict[str, Any] | None:
+        try:
+            return self._read(run_id)
+        except ReleaseRunNotFound:
+            return None
+
     def _assert_workspace(self, record: dict[str, Any]) -> None:
         if (
             record["immutable"]["input"]["workspace_fingerprint"]
@@ -577,19 +665,48 @@ class ReleaseRunStore:
                     "Release run record permissions could not be secured."
                 )
             _fsync_directory(self.root)
+        except ReleaseRunError:
+            raise
+        except OSError as exc:
+            raise ReleaseRunCorrupt(
+                "Release run record could not be persisted safely."
+            ) from exc
         finally:
             if descriptor >= 0:
-                os.close(descriptor)
+                try:
+                    os.close(descriptor)
+                except OSError as exc:
+                    raise ReleaseRunCorrupt(
+                        "Release run record could not be persisted safely."
+                    ) from exc
             if temporary_path is not None:
                 try:
                     os.unlink(temporary_path)
-                except FileNotFoundError:
+                except OSError:
                     pass
 
     def _locked(self):  # type: ignore[no-untyped-def]
         self._ensure_root()
+        self._validate_lock_path()
         return _CombinedLock(self._thread_lock, FileLock(str(self._lock_path)))
 
+    def _validate_lock_path(self) -> None:
+        try:
+            metadata = self._lock_path.lstat()
+        except FileNotFoundError:
+            return
+        except OSError as exc:
+            raise ReleaseRunCorrupt(
+                "Release run lock could not be inspected safely."
+            ) from exc
+        if (
+            stat.S_ISLNK(metadata.st_mode)
+            or not stat.S_ISREG(metadata.st_mode)
+            or metadata.st_nlink != 1
+            or (hasattr(os, "geteuid") and metadata.st_uid != os.geteuid())
+        ):
+            raise ReleaseRunCorrupt("Release run lock path is unsafe.")
+
 
 class _CombinedLock:
     def __init__(self, thread_lock: threading.RLock, file_lock: FileLock) -> None:
@@ -605,13 +722,23 @@ class _CombinedLock:
             raise ReleaseRunConflict(
                 "Release run storage is busy; retry the operation."
             ) from exc
+        except OSError as exc:
+            self.thread_lock.release()
+            raise ReleaseRunCorrupt(
+                "Release run lock could not be acquired safely."
+            ) from exc
         except BaseException:
             self.thread_lock.release()
             raise
 
     def __exit__(self, *exc_info: object) -> None:
         try:
-            self.file_lock.release()
+            try:
+                self.file_lock.release()
+            except OSError as exc:
+                raise ReleaseRunCorrupt(
+                    "Release run lock could not be released safely."
+                ) from exc
         finally:
             self.thread_lock.release()
 
@@ -755,6 +882,7 @@ def _validate_record(value: Any, *, expected_run_id: str | None = None) -> None:
             "schema",
             "schema_version",
             "run_id",
+            "create_request_fingerprint",
             "created_at",
             "updated_at",
             "immutable",
@@ -773,6 +901,10 @@ def _validate_record(value: Any, *, expected_run_id: str | None = None) -> None:
             raise ValueError
         if expected_run_id is not None and run_id != expected_run_id:
             raise ValueError
+        if not _is_sha256(value.get("create_request_fingerprint")):
+            raise ValueError
+        if run_id != _creation_run_id(value["create_request_fingerprint"]):
+            raise ValueError
         for key in ("created_at", "updated_at"):
             if not _is_timestamp(value.get(key)):
                 raise ValueError
@@ -908,8 +1040,15 @@ def _validate_record(value: Any, *, expected_run_id: str | None = None) -> None:
             if (
                 not isinstance(request, dict)
                 or set(request)
-                != {"fingerprint", "operation", "step_id", "parameter"}
+                != {
+                    "fingerprint",
+                    "operation",
+                    "step_id",
+                    "parameter",
+                    "attempt_fingerprint",
+                }
                 or not _is_sha256(request.get("fingerprint"))
+                or not _is_sha256(request.get("attempt_fingerprint"))
                 or request.get("operation")
                 not in {"resume", "retry", "reconcile"}
             ):
@@ -961,6 +1100,29 @@ def _validate_record(value: Any, *, expected_run_id: str | None = None) -> None:
         attempts_by_fingerprint = {
             attempt["fingerprint"]: attempt for attempt in attempts
         }
+        for request in requests:
+            request_attempt = attempts_by_fingerprint.get(
+                request["attempt_fingerprint"]
+            )
+            if request_attempt is None:
+                raise ValueError
+            if (
+                request["step_id"] is not None
+                and request_attempt["step_id"] != request["step_id"]
+            ):
+                raise ValueError
+            if request["operation"] == "resume" and request_attempt[
+                "outcome"
+            ] != "interrupted":
+                raise ValueError
+            if request["operation"] == "retry" and request_attempt[
+                "outcome"
+            ] not in {"failed", "interrupted"}:
+                raise ValueError
+            if request["operation"] == "reconcile" and request_attempt[
+                "outcome"
+            ] != "interrupted":
+                raise ValueError
         attempts_per_step = {
             step_id: sum(
                 attempt["step_id"] == step_id for attempt in attempts
@@ -1193,13 +1355,28 @@ def _append_event(
     record["state"]["events"] = record["state"]["events"][-MAX_EVENTS:]
 
 
-def _ensure_command_capacity(record: dict[str, Any]) -> None:
-    if len(record["state"]["processed_requests"]) >= MAX_COMMANDS:
+def _ensure_command_capacity(
+    record: dict[str, Any], *, reserve_after: int = 0
+) -> None:
+    if (
+        len(record["state"]["processed_requests"]) + 1 + reserve_after
+        > MAX_COMMANDS
+    ):
         raise ReleaseRunConflict(
             "Release run command history reached its safety limit; create a new run."
         )
 
 
+def _ensure_start_recovery_capacity(record: dict[str, Any]) -> None:
+    if (
+        len(record["state"]["processed_requests"]) + START_RECOVERY_RESERVE
+        > MAX_COMMANDS
+    ):
+        raise ReleaseRunConflict(
+            "Release run has insufficient reserved recovery capacity for a new attempt; create a new run."
+        )
+
+
 def _ensure_attempt_capacity(record: dict[str, Any]) -> None:
     if len(record["state"]["processed_attempts"]) >= MAX_COMMANDS:
         raise ReleaseRunConflict(
@@ -1213,6 +1390,8 @@ def _remember_request(
     operation: str,
     step_id: str | None,
     parameter: str | None,
+    *,
+    attempt_fingerprint: str | None,
 ) -> None:
     _ensure_command_capacity(record)
     record["state"]["processed_requests"].append(
@@ -1221,6 +1400,7 @@ def _remember_request(
             "operation": operation,
             "step_id": step_id,
             "parameter": parameter,
+            "attempt_fingerprint": attempt_fingerprint,
         }
     )
 
@@ -1324,6 +1504,28 @@ def _record_digest(record: dict[str, Any]) -> str:
     ).hexdigest()
 
 
+def _creation_run_id(request_fingerprint: str) -> str:
+    if not _is_sha256(request_fingerprint):
+        raise ReleaseRunConflict("Run creation request identifier is invalid.")
+    return f"rr-request-{request_fingerprint}"
+
+
+def _validate_create_replay(
+    record: dict[str, Any],
+    *,
+    request_fingerprint: str,
+    input_snapshot: dict[str, Any],
+) -> None:
+    if record["create_request_fingerprint"] != request_fingerprint:
+        raise ReleaseRunConflict(
+            "The run creation identifier is already mapped to another record."
+        )
+    if record["immutable"]["input"] != input_snapshot:
+        raise ReleaseRunConflict(
+            "The run creation identifier was already used with different inputs."
+        )
+
+
 def _request_fingerprint(request_id: str) -> str:
     if not isinstance(request_id, str) or not REQUEST_ID_PATTERN.fullmatch(request_id):
         raise ReleaseRunConflict("Run-state request identifier is invalid.")
diff --git a/tools/release/server/app.py b/tools/release/server/app.py
index 4f70da1..cfde3c5 100644
--- a/tools/release/server/app.py
+++ b/tools/release/server/app.py
@@ -100,6 +100,7 @@ class RepositoryTagRequest(BaseModel):
 
 
 class ReleaseRunCreateRequest(BaseModel):
+    request_id: str = Field(min_length=8, max_length=128)
     repo_versions: dict[str, str] = Field(default_factory=dict)
     channel: str = Field(default="stable", min_length=1, max_length=64)
     online: bool = False
@@ -128,10 +129,15 @@ def create_app(
     app.state.workspace_root = workspace_root
     app.state.token = token
     app.state.workspace_fingerprint = release_workspace_fingerprint(workspace_root)
-    app.state.release_runs = ReleaseRunStore(
-        run_state_root
+    release_run_root = (
+        workspace_release_run_root(
+            run_state_root, app.state.workspace_fingerprint
+        )
         if run_state_root is not None
-        else default_release_run_root(workspace_root),
+        else default_release_run_root(workspace_root)
+    )
+    app.state.release_runs = ReleaseRunStore(
+        release_run_root,
         expected_workspace_fingerprint=app.state.workspace_fingerprint,
     )
 
@@ -255,6 +261,24 @@ def create_app(
             raise HTTPException(
                 status_code=400, detail="At least one repository version is required"
             )
+        input_snapshot = {
+            "channel": request.channel,
+            "repo_versions": dict(request.repo_versions),
+            "online": request.online,
+            "remote_tags": request.remote_tags,
+            "public_catalog": request.public_catalog,
+            "include_migrations": request.include_migrations,
+            "workspace_fingerprint": app.state.workspace_fingerprint,
+        }
+        try:
+            existing = app.state.release_runs.find_created_run(
+                request_id=request.request_id,
+                input_snapshot=input_snapshot,
+            )
+        except (ReleaseRunConflict, ReleaseRunCorrupt) as exc:
+            raise HTTPException(status_code=409, detail=str(exc)) from exc
+        if existing is not None:
+            return existing
         snapshot = build_dashboard(
             workspace_root=app.state.workspace_root,
             online=request.online,
@@ -294,15 +318,8 @@ def create_app(
             )
         try:
             return app.state.release_runs.create(
-                input_snapshot={
-                    "channel": request.channel,
-                    "repo_versions": dict(request.repo_versions),
-                    "online": request.online,
-                    "remote_tags": request.remote_tags,
-                    "public_catalog": request.public_catalog,
-                    "include_migrations": request.include_migrations,
-                    "workspace_fingerprint": app.state.workspace_fingerprint,
-                },
+                request_id=request.request_id,
+                input_snapshot=input_snapshot,
                 plan_snapshot=plan_payload,
             )
         except (ReleaseRunConflict, ReleaseRunCorrupt) as exc:
@@ -492,6 +509,14 @@ def default_release_run_root(workspace_root: Path) -> Path:
     )
 
 
+def workspace_release_run_root(state_root: Path, workspace_fingerprint: str) -> Path:
+    return (
+        state_root.expanduser()
+        / f"workspace-{workspace_fingerprint}"
+        / "release-runs"
+    )
+
+
 def release_workspace_fingerprint(workspace_root: Path) -> str:
     canonical = str(workspace_root.expanduser().resolve()).encode("utf-8")
     return hashlib.sha256(canonical).hexdigest()
diff --git a/tools/release/webui/index.html b/tools/release/webui/index.html
index c95e74e..f27b54a 100644
--- a/tools/release/webui/index.html
+++ b/tools/release/webui/index.html
@@ -909,6 +909,8 @@
         runs: [],
         runStoreAvailable: true,
       };
+      const pendingReleaseRunKey = "govoplanReleaseConsolePendingRunCreate";
+      const pendingRunCommandsKey = "govoplanReleaseConsolePendingRunCommands";
 
       elements.refresh.addEventListener("click", load);
       elements.generateCandidate.addEventListener("click", generateCandidate);
@@ -962,7 +964,9 @@
         }).then((response) => {
           if (!response.ok) {
             return response.json().catch(() => ({})).then((payload) => {
-              throw new Error(payload.detail || `${response.status} ${response.statusText}`);
+              const error = new Error(payload.detail || `${response.status} ${response.statusText}`);
+              error.status = response.status;
+              throw error;
             });
           }
           return response.json();
@@ -980,7 +984,9 @@
         }).then((response) => {
           if (!response.ok) {
             return response.json().catch(() => ({})).then((payload) => {
-              throw new Error(payload.detail || `${response.status} ${response.statusText}`);
+              const error = new Error(payload.detail || `${response.status} ${response.statusText}`);
+              error.status = response.status;
+              throw error;
             });
           }
           return response.json();
@@ -1000,9 +1006,10 @@
             renderReleaseRunStoreUnavailable(runsResult.error);
           } else {
             renderReleaseRunList(runsResult.data?.runs || []);
-            if (releaseState.currentRun?.run_id) {
+            const pendingRecoveryResult = await recoverPendingReleaseRun();
+            if (pendingRecoveryResult === null && releaseState.currentRun?.run_id) {
               await loadReleaseRun(releaseState.currentRun.run_id, { restoreSelection: false });
-            } else {
+            } else if (pendingRecoveryResult === null) {
               renderIdlePlan();
             }
           }
@@ -1070,7 +1077,7 @@
         setBusy([elements.createReleaseRun], true);
         elements.runStatus.textContent = "freezing plan...";
         try {
-          const run = await postJson("/api/release-runs", {
+          const payload = pendingReleaseRunPayload({
             repo_versions: repoVersions,
             channel: elements.channel.value.trim() || "stable",
             online: false,
@@ -1078,12 +1085,7 @@
             public_catalog: elements.publicCatalog.checked,
             include_migrations: elements.migrations.checked,
           });
-          releaseState.currentRun = run;
-          restoreReleaseRunSelection(run);
-          renderReleaseRun(run);
-          const runs = await api("/api/release-runs");
-          renderReleaseRunList(runs.runs || []);
-          elements.releaseRun.value = run.run_id;
+          await submitReleaseRunCreate(payload, { automatic: false });
         } catch (error) {
           elements.runStatus.textContent = "error";
           elements.runOutput.innerHTML = `<div class="action"><h3>${pill("error", "block")} Run creation failed</h3><p>${escapeHtml(error.message)}</p></div>`;
@@ -1092,6 +1094,74 @@
         }
       }
 
+      function pendingReleaseRunPayload(input) {
+        const normalized = {
+          ...input,
+          repo_versions: Object.fromEntries(Object.entries(input.repo_versions || {}).sort(([left], [right]) => left.localeCompare(right))),
+        };
+        const signature = JSON.stringify(normalized);
+        const pending = readPendingReleaseRun();
+        if (pending?.signature === signature) return pending.payload;
+        const payload = { ...normalized, request_id: requestId("create-run") };
+        sessionStorage.setItem(pendingReleaseRunKey, JSON.stringify({ signature, payload }));
+        return payload;
+      }
+
+      function readPendingReleaseRun() {
+        try {
+          const pending = JSON.parse(sessionStorage.getItem(pendingReleaseRunKey) || "null");
+          if (!pending || typeof pending.signature !== "string" || typeof pending.payload?.request_id !== "string") return null;
+          return pending;
+        } catch (_error) {
+          sessionStorage.removeItem(pendingReleaseRunKey);
+          return null;
+        }
+      }
+
+      function clearPendingReleaseRun(requestIdToClear) {
+        const pending = readPendingReleaseRun();
+        if (pending?.payload?.request_id === requestIdToClear) {
+          sessionStorage.removeItem(pendingReleaseRunKey);
+        }
+      }
+
+      async function recoverPendingReleaseRun() {
+        const pending = readPendingReleaseRun();
+        if (!pending) return null;
+        if (!releaseState.runStoreAvailable) return false;
+        elements.runStatus.textContent = "recovering saved create...";
+        return submitReleaseRunCreate(pending.payload, { automatic: true });
+      }
+
+      async function submitReleaseRunCreate(payload, options = {}) {
+        try {
+          const run = await postJson("/api/release-runs", payload);
+          clearPendingReleaseRun(payload.request_id);
+          releaseState.currentRun = run;
+          restoreReleaseRunSelection(run);
+          renderReleaseRun(run);
+          await refreshReleaseRunListAfterCreate(run);
+          return true;
+        } catch (error) {
+          const deterministicClientError = error.status >= 400 && error.status < 500 && error.status !== 409;
+          const conflictingReuse = error.status === 409 && error.message.includes("already used with different inputs");
+          if (deterministicClientError || conflictingReuse) clearPendingReleaseRun(payload.request_id);
+          elements.runStatus.textContent = options.automatic ? "recovery pending" : "error";
+          elements.runOutput.innerHTML = `<div class="action"><h3>${pill(options.automatic ? "pending" : "error", options.automatic ? "warn" : "block")} ${options.automatic ? "Saved run recovery is still pending" : "Run creation failed"}</h3><p>${escapeHtml(error.message)}</p><p class="action-meta">The same private request identifier will be reused after an uncertain response.</p></div>`;
+          return false;
+        }
+      }
+
+      async function refreshReleaseRunListAfterCreate(run) {
+        try {
+          const runs = await api("/api/release-runs");
+          renderReleaseRunList(runs.runs || []);
+          elements.releaseRun.value = run.run_id;
+        } catch (error) {
+          elements.runOutput.insertAdjacentHTML("afterbegin", `<div class="action"><h3>${pill("saved", "warn")} Run saved; list refresh unavailable</h3><p>${escapeHtml(error.message)}</p><p class="action-meta">The persisted run remains selected and can be reloaded after refresh.</p></div>`);
+        }
+      }
+
       async function loadReleaseRun(runId, options = {}) {
         if (!runId) {
           releaseState.currentRun = null;
@@ -1183,9 +1253,12 @@
       async function resumeReleaseRun() {
         const run = releaseState.currentRun;
         if (!run) return;
+        const commandKey = `resume:${run.run_id}`;
+        const commandRequestId = inFlightRunCommandId(commandKey, "resume");
         setBusy([elements.resumeReleaseRun], true);
         try {
-          const resumed = await postJson(`/api/release-runs/${encodeURIComponent(run.run_id)}/resume`, { request_id: requestId("resume") });
+          const resumed = await postJson(`/api/release-runs/${encodeURIComponent(run.run_id)}/resume`, { request_id: commandRequestId });
+          clearInFlightRunCommand(commandKey, commandRequestId);
           releaseState.currentRun = resumed;
           renderReleaseRun(resumed);
         } catch (error) {
@@ -1199,8 +1272,11 @@
       async function retryReleaseRunStep(stepId) {
         const run = releaseState.currentRun;
         if (!run || !stepId) return;
+        const commandKey = `retry:${run.run_id}:${stepId}`;
+        const commandRequestId = inFlightRunCommandId(commandKey, "retry");
         try {
-          const retried = await postJson(`/api/release-runs/${encodeURIComponent(run.run_id)}/steps/${encodeURIComponent(stepId)}/retry`, { request_id: requestId("retry") });
+          const retried = await postJson(`/api/release-runs/${encodeURIComponent(run.run_id)}/steps/${encodeURIComponent(stepId)}/retry`, { request_id: commandRequestId });
+          clearInFlightRunCommand(commandKey, commandRequestId);
           releaseState.currentRun = retried;
           renderReleaseRun(retried);
         } catch (error) {
@@ -1211,13 +1287,16 @@
       async function reconcileReleaseRunStep(stepId, outcome, confirmation, button) {
         const run = releaseState.currentRun;
         if (!run || !stepId || !outcome || confirmation !== "RECONCILE") return;
+        const commandKey = `reconcile:${run.run_id}:${stepId}:${outcome}`;
+        const commandRequestId = inFlightRunCommandId(commandKey, "reconcile");
         setBusy([button], true);
         try {
           const reconciled = await postJson(`/api/release-runs/${encodeURIComponent(run.run_id)}/steps/${encodeURIComponent(stepId)}/reconcile`, {
-            request_id: requestId("reconcile"),
+            request_id: commandRequestId,
             outcome,
             confirm: confirmation,
           });
+          clearInFlightRunCommand(commandKey, commandRequestId);
           releaseState.currentRun = reconciled;
           renderReleaseRun(reconciled);
         } catch (error) {
@@ -1226,6 +1305,33 @@
         }
       }
 
+      function readInFlightRunCommands() {
+        try {
+          const commands = JSON.parse(sessionStorage.getItem(pendingRunCommandsKey) || "{}");
+          return commands && typeof commands === "object" && !Array.isArray(commands) ? commands : {};
+        } catch (_error) {
+          sessionStorage.removeItem(pendingRunCommandsKey);
+          return {};
+        }
+      }
+
+      function inFlightRunCommandId(commandKey, prefix) {
+        const commands = readInFlightRunCommands();
+        if (typeof commands[commandKey] === "string") return commands[commandKey];
+        const identifier = requestId(prefix);
+        commands[commandKey] = identifier;
+        sessionStorage.setItem(pendingRunCommandsKey, JSON.stringify(commands));
+        return identifier;
+      }
+
+      function clearInFlightRunCommand(commandKey, requestIdToClear) {
+        const commands = readInFlightRunCommands();
+        if (commands[commandKey] !== requestIdToClear) return;
+        delete commands[commandKey];
+        if (Object.keys(commands).length) sessionStorage.setItem(pendingRunCommandsKey, JSON.stringify(commands));
+        else sessionStorage.removeItem(pendingRunCommandsKey);
+      }
+
       function requestId(prefix) {
         const identifier = window.crypto?.randomUUID ? window.crypto.randomUUID() : `${Date.now()}-${Math.random().toString(16).slice(2)}`;
         return `${prefix}-${identifier}`;