[S.J. (cstolting-collab)]

A receive-side link handler failure can bypass incoming-window replenishment on #49368. The asynchronous unattached-handle path can also advertise receive capacity during teardown. This draft adds the incoming correction directly to EldertGrootenboer's branch while preserving its outgoing-credit, fragmentation, and concurrency implementation.

Scope and relationship

• Base: fix/pyamqp-session-window-49232 at 8c13500388c7330820d53a7bac27c116cfedd339.
• One commit; seven changed files; 434 insertions and 14 deletions, verified against that actual base before opening.
• Four mirrored sync/async session modules, two incoming-window regression modules, and one Chronus entry.
• Recovery runs after dispatch in finally, only while the session is MAPPED and incoming_window <= 0.
• Replaces the oversized review packaging in [Follow-up to #49368] Recover incoming pyamqp windows and resolve changelogs #49947. This branch does not include that PR's merge of newer main or its two changelog conflict resolutions. Main synchronization remains a separate owner integration decision.
• Service Bus/Event Hubs: prevent pyamqp session-window exhaustion #49270 carries the same incoming correction with a different outgoing policy. This draft offers the incoming fix on top of fix: Prevent pyAMQP session window exhaustion #49368; the two contributions should not be independently merged as separate implementations of the same fix. Service Bus/Event Hubs: prevent pyamqp session-window exhaustion #49270 remains unchanged for maintainer consideration.

Validation

The November 11 inspection independently reran the focused unit selection:

• On fix: Prevent pyAMQP session window exhaustion #49368 baseline 8c13500388c7330820d53a7bac27c116cfedd339, the new 54-case selection produced 36 failures and 18 passes.
• On the incoming-only source snapshot, 386 focused tests passed: 332 existing session-window cases plus the 54 incoming cases.
• The new cases cover exhausted/negative windows, handler errors/cancellation, unattached handles, teardown, positive remaining capacity, and dispatch before FLOW failure/cancellation.
• All seven correction files are byte-identical to the correction in [Follow-up to #49368] Recover incoming pyamqp windows and resolve changelogs #49947. The two new test modules are also byte-identical to those in Service Bus/Event Hubs: prevent pyamqp session-window exhaustion #49270.
• The four session modules are unchanged from fix: Prevent pyAMQP session window exhaustion #49368 outside _incoming_transfer; its other implementation files remain unchanged.
• The seven-file diff passes git diff --check.

Published commit: c57983c591ca560a16d417a723da8f58485edc6f.
Exact tested/published Git tree: e58e5e8d910e3a9081dcfd637284e204b5f79549.
Local pre-merge correction commit: a84cd076286445adffbc3a306ee35dffdd5514dd.
Publication recreated the commit through GitHub with the same complete tree and the same sole parent; no existing branch history was rewritten.

Tests used Linux/Python 3.12.14 and checked-out package sources via PYTHONPATH, with --noconftest --import-mode=importlib. This is a focused pure-unit selection with mocks, not complete package or live-broker validation. Handler and FLOW failures are exercised separately; no simultaneous-failure exception-priority guarantee is asserted. Historical lint results are retained with #49947's evidence and were not rerun for this packaging-only publication.

CI and handoff

Remote build/test validation of this new commit is pending. The relevant automatic pipeline branch filters do not include the fix/* target, so successful local tests must not be treated as successful remote CI; an authorized maintainer may need to arrange a run.

#49947 and its original branch/evidence are preserved as the historical integration attempt. This draft is the focused review vehicle for the incoming-only correction.

Contribution checklist

• No public API or intended breaking changes.
• Chronus change entry supplied for both packages; main/changelog synchronization is separate.
• Contribution guidance reviewed; one focused commit with regression coverage.

[azure-sdk-automation]

The sdk-api-review-ready label cannot create a review PR for this working PR because its branch is in a fork repository. Prefer a working branch in the registered upstream repository, or reach out to the ARH maintainers for help. Fork reviews are supported through the Azure SDK CLI (azsdk apireview create), but the fork requires its own ARH registration and the review PR is created in the fork rather than the upstream repository.

[azure-pipelines]

Azure Pipelines:
8 pipeline(s) were filtered out due to trigger conditions.
There may be pipelines that require an authorized user to comment /azp run to run.

[github-actions]

Thank you for your contribution S.J. (@cstolting-collab)! We will review the pull request and get back to you soon.