Approval and observation are different events
Preserve the decision to approve and the later report of loaded state as separate records.
Research note · November 14, 2026Research question
Why can an approved release differ from a runtime observation without either record being internally inconsistent?
Method
- Approve a new fixture release without restarting the worker. Record the approval time, destination and artifact identities independently of the worker’s latest report.
- Collect an observation before the reload and another after an explicit reload. Keep the ordering and observation source visible; do not overwrite the earlier state when a later report arrives.
- Classify approval, reload request, reload completion and observed identity as separate events. A missing event should remain missing rather than being inferred from the next available status.
Records to retain
- Approval decision and target
- Reload request identifier
- Before and after reports
- Event and collection timestamps
Source basis
OpenTelemetry separates event time from observed time, and W3C PROV distinguishes activities from the entities they use or produce. These distinctions support a release history that does not equate a decision with execution.
Interpretation & limits
The protocol depends on the report’s trust boundary. A worker self-report is not independent proof of loaded memory. It does not demonstrate production deployment control.
Next experiments
- Add an approval that is never followed by a reload.
- Test whether the UI preserves both facts during a mismatch.
References
Technical reference
Implementation details supporting this note, drawn directly from the maintained documentation. These are technical references, not additional papers or new experimental results. Expand a topic to inspect its method, fields and limits.