[soulteary]

Problem

A physically sent angle command can retain Dispatched status after its verified response is missing. Under the explicit Python policy, C++ also recorded that operation as a successful logical transaction and cleared its consecutive-error count. The Python SDK retains physical dispatch while recording a response timeout and leaving successful latency unchanged.

A post-merge release audit reproduced the discrepancy in the v0.3.9 source: dropping one command response increased C++'s successful transaction count by one; the reference Python count did not increase. Public release was held. The signed v0.3.9 tag remains unchanged and was not published as a Release.

Change

At optional angle-feedback expiry, mark the Python logical timing scope unsuccessful and record one response-timeout error. Retain the command's physical send result and terminal Dispatched status. A subsequent complete logical transaction clears consecutive errors. Native optional-feedback accounting is unchanged. L20 speed and current-limit writes retain their dispatch-only Python contract; missing native optional observations do not create logical errors.

Released v0.3.10, including the health corrections from #23. Existing method signatures, Native defaults and the transport v0.3.4 minimum remain unchanged. Rebuild the SDK and consumers because the cumulative update from v0.3.8 includes appended public structure fields; binary compatibility with older headers is not claimed.

Validation

• Four added regressions cover consecutive missing feedback and later success, missing feedback without polling, unchanged Native accounting, and L20 non-angle optional observations. The complete health regression selection passes 18 cases / 177 assertions.

• 59 reference Python runtime tests passed, including the missing-feedback delivery contract. Reference source: 3a63e723b9a0046e9a60a58339b9f064ef5b70fd; this is not a claimed deployment-build SHA.

• Documentation and validation-helper checks passed. All 10 local CTest groups and the installed-package consumer passed. The same installed-SDK reproduction now retains the successful count (4 → 4); 6144 percentile-window updates still match the Python oracle and actual reference RuntimeMetrics. Final-head checks passed:

• CI: GCC Debug/Release, Windows x64 Debug/Release and ASan/UBSan.

• Released transport consumers: x86_64 and ARM64 with transport v0.3.4.

• Complete static analysis: downloaded report independently checked; A+, score 100, zero diagnostics; 619/619 formatting files and 362/362 units for each analyzer.

• Release acceptance: all twelve native targets and complete package collection.

No new hardware acceptance is claimed. Adapter/driver and firmware combinations, physical timestamps, mixed CAN/FD topology and extended multi-device operation require separate hardware validation. O20, Python bindings and hand/arm integration are excluded.

Released source verification

Merge and signed tag: 5aa263c. The merged tree matches the tested PR head.

• Merged-source CI
• Released-transport consumers
• Complete A+ report
• Twelve-target release acceptance and verified draft

All merged-source gates and all 15 manual release jobs passed. The merged full report was independently checked: A+, score 100, zero diagnostics; 619/619 formatting files and 362/362 units per analyzer. All 26 downloaded draft assets matched API SHA-256 digests and sizes; package source identities, object architectures and archive checksums were revalidated. Public stable/latest status, signed tag, release notes and all published asset bytes were checked after publication.

An initial merged-source Windows x64 Release run timed out in the existing Native L20 current-limit fixture. The selected job passed when rerun on the unchanged source, assertions and timeout settings. The first failure remains in the workflow history; its root cause was not established.