The runtime reload boundary
Observe the transition rather than treating a reload request as success.
Research note · November 14, 2026Research question
Which evidence is needed to connect a requested update to a subsequent loaded-state report?
Method
- Assign a request identifier to a controlled reload operation. Record the old and requested release identities before invoking it.
- Collect acceptance, completion or failure information independently. After completion, collect a fresh runtime report and compare its identity with the intended release.
- Exercise a failed reload and an interrupted collector. Preserve a last-known state without inventing a successful transition when the final observation is unavailable.
Records to retain
- Reload request and target digest
- Operation outcome
- Latest pre-reload observation
- First valid post-reload observation
Source basis
HTTP semantics distinguishes accepting a request from the work that a system may perform afterward. Provenance modeling similarly separates an activity from its resulting entity. A release transition requires evidence at both boundaries.
Interpretation & limits
A process can report a desired identity incorrectly. This is a test of observation sequencing, not hardware-backed proof or automatic repair. It proposes no public Contour control API.
Next experiments
- Inject a load failure into a local fixture.
- Check that an accepted request cannot directly produce a matching status.
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.