Prepare Python 2.5.15 query authority without regressing backfill docs

CHANGELOG.md

@@ -6,6 +6,22 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [2.5.15] - 2026-10-15
+
+### Fixed
+- Bound query payload loading, replay and async handler execution to the
+  original query deadline and lease. Use non-renewing authority observations
+  when Server 2.8.0 advertises them, with explicit older-Server/1.19 limits.
+- Reuse identical external references within one query response and retain
+  the original typed caller timeout. Add optional pre-registration enforcement
+  with `require_query_task_authority=True`.
+
+## [2.5.14] - 2026-10-14
+
+### Fixed
+- Document schedule backfill as a start-inclusive, end-exclusive window.
+  An occurrence at the end belongs to the next adjacent window.
+
 ## [2.5.13] - 2026-10-13
 
 ### Fixed

docs/reference/client.md

@@ -6,4 +6,28 @@ Workflow authors should normally yield `ctx.call_nexus_service(...)`; the
 worker uses this client method to execute once and record the durable result
 or typed service failure.
 
+## Backfill windows
+
+`Client.backfill_schedule()` includes the start time and excludes the end time:
+`[start_time, end_time)`. An occurrence exactly at `end_time` is excluded.
+
+Adjacent windows share a boundary without overlapping it:
+
+```python
+await client.backfill_schedule(
+    "daily-orders",
+    start_time="2026-03-08T07:00:00Z",
+    end_time="2026-03-08T07:30:00Z",
+)
+await client.backfill_schedule(
+    "daily-orders",
+    start_time="2026-03-08T07:30:00Z",
+    end_time="2026-03-08T08:00:00Z",
+)
+```
+
+An occurrence at `07:30:00Z` belongs to the second window. Repeating a window
+does not start another workflow for an occurrence already admitted by that
+schedule.
+
 ::: durable_workflow.client

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "durable-workflow"
-version = "2.5.14"
+version = "2.5.15"
 description = "Python client and worker SDK for Durable Workflow Cloud and self-hosted Server"
 readme = "README.md"
 requires-python = ">=3.10"
@@ -71,8 +71,8 @@ durable-workflow-replay-conformance = "durable_workflow.replay_conformance:main"
 durable-workflow-workflow-updates-conformance = "durable_workflow.workflow_updates_conformance:main"
 
 [tool.durable-workflow]
-product-train = "2.5.14"
-registry-version = "2.5.14"
+product-train = "2.5.15"
+registry-version = "2.5.15"
 supported-server-versions = "2.8.0"
 worker-protocol-version = "1.20"
 control-plane-version = "2"

src/durable_workflow/client.py

@@ -4845,8 +4845,10 @@ async def backfill_schedule(
         end_time: str,
         overlap_policy: str | None = None,
     ) -> ScheduleBackfillResult:
-        """Fire a schedule for every would-have-been moment in ``[start_time, end_time]``.
+        """Fire a schedule for every would-have-been moment in ``[start_time, end_time)``.
 
+        The start is inclusive and the end is exclusive. An occurrence equal
+        to ``end_time`` is excluded and belongs to the next adjacent window.
         Times are ISO-8601 strings. Useful to replay a period the schedule
         was paused or to seed historical runs. The returned
         :class:`ScheduleBackfillResult` reports how many fires were attempted