Refresh and remove one disposable cache through separate governed runs

This alpha profile prepares closed local inputs, runs a bounded synthetic cache refresh, inspects the retained result, obtains a fresh observation, and runs a separately authorized teardown. It is a development qualification on a caller-owned disposable Docker host.

What participates

closed Maude plan → Nightshift observation/proposal → AG decision
  → Docket custody → disposable Docker executor → retained NQ result
  → fresh same-family observation → separate AG/Docket teardown

NQ admits the exact recorded result for this run. Pulse supplies the named support adapter. Maude validates and locks the plan but does not accept a proposal or grant permission. The selected Docker daemon remains a deployment trust boundary.

Public source pins

Clone clean detached checkouts from the six public repositories at the exact commits in manifest.json. Use Linux x86-64, Rust/Cargo 1.94.0, CPython 3.12.3, Docker client/server 29.1.3 and Compose 5.0.0. Build each required binary with its locked dependency graph, copy the regular executable files into a caller-owned program directory, and retain their SHA-256 values. Build capacity and the required local host reserve are caller responsibilities.

Download or clone this release snapshot, verify the release-local input helper, and use it with an absent caller-owned output directory. The helper accepts only the pinned, clean source cohort and measured regular files. The synthetic runtime fixture root passed to the Maude driver must be a fresh path below /tmp. Before launch, choose a separate durable manager/control and retained-custody path outside /tmp; the fixture location is not durable evidence storage. Then follow the pinned Maude connected-cache prerequisite guide and invoke its run_connected_cache.py --execute driver under a durable user manager.

Inspect before another submission

After the manager stops, inspect the original terminal record, AG decision/settlement, Docket attempt, exact NQ result and Docker project labels using the pinned Maude operations guide. If a terminal record, process state, or owner result is missing, preserve the original root and report the outcome as indeterminate. Do not reuse an existing root or repeat a possibly uncertain occurrence.

Verified scope

The clean public-only newcomer case completed 53 retained stages, two exact governed occurrences, NQ admission of the recorded result, a fresh successor observation and project absence after separate teardown. Its copied existing-root invocation refused before any later stage or Docker call. No provider call or automatic retry occurred. Two earlier recovery checks refused before effects: one rejected an unsafe NQ helper ancestor and one rejected synthetic evidence outside /tmp. A separate case stopped the supervisor after both settlements and retained reconciliation evidence. That case does not establish general restart/resume, interrupted-action recovery, or automatic retry.

No provider credential is needed to replay the deterministic path. Live notification delivery, external review dispatch, a production cache, continuous operation, federation, ECAD/design-flow readiness, restore/rollback and cross-version migration are outside this alpha. See qualification.json, release notes, operations and troubleshooting.