[trissim]

A terminal STATUS reply can arrive before queued progress callbacks finish. In the retained OpenHCS Speckles run, execution and all three CSV exports succeeded, but disconnect discarded the final step/export/axis notifications and the worker-coverage check failed.

ExecutionRecord now owns the sequence of admitted progress and its latest detached event. Registration establishes a producer baseline before the listener starts. Each published event carries that sequence, and wait_for_completion waits on the existing subscriber condition until callbacks finish through the terminal record’s watermark. FAILED and CANCELLED receipts retain their actual record too. Socket polling wakes immediately for available events while keeping listener shutdown bounded.

Late subscribers start at the registration baseline without fabricated historical callbacks. Missing tails time out; sequence gaps and callback failures produce explicit delivery errors, preserved across reconnects. This detects PUB loss; it does not add a journal or guarantee replay, and it does not change the server execution outcome. No additional STATUS request, sleep-based completion delay, or coverage waiver is introduced.

Validation: 79 passing tests across test_execution.py, test_progress_stream.py, and test_messages.py, including blocked callback tails for COMPLETE/FAILED/CANCELLED, late registration, missing/gapped events, callback failure, reconnect error retention, and existing listener ownership. git diff --check passes. Existing repository lint debt remains outside this change. The original scientific workload has not been rerun with this patch yet.

Closes #18
Refs OpenHCSDev/openhcs#573
Refs OpenHCSDev/openhcs#394

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.