research(track-a): prove runtime reacquisition stability - #303
Conversation
|
Coordinator disposition at current head The namespace/credential/cleanup design is materially improved and standard CI for code-bearing head Required evidence before reconsideration:
Until those execute, restart/relogin stability, bridge session lifecycle and R4 semantics remain UNKNOWN. Historical one-off movement or process survival does not satisfy this gate. |
|
Coordinator revalidation input (PR #300): the prior “newer child required” prerequisite is superseded by exact evidence. FACT:
BOUNDARY:
Durable coordinator evidence: |
|
Coordinator disposition: CLOSE_SUPERSEDED / historical evidence only. Current trusted programme state has moved to the canonical persistent-runtime admission model and the sole current physical RUNTIME lane is PR #358. #358's fresh read-only reconciliation on PR #303 is based on stale The branch and its evidence remain preserved for historical/non-authoritative research reference. Closing this PR does not reject previously recorded experimental observations; it prevents stale workflow/runtime ownership from being resumed or mistaken for the canonical session. Any still-useful bounded evidence must be explicitly revalidated/promoted by the coordinator under current main. No physical cleanup or runtime mutation is authorized by this closure. |
Draft-only Track A runtime research dispatch from exact
main@8fca1c3eee453d0d4ef8a47e0f15c9dbae491b45.Hypothesis: after clean task-owned restart/relogin, the exact-build structural world observer/read path can be reacquired from fresh PID/PIE/runtime state.
The task incorporates the #289 safety corrections: credentials must not be job-scoped into persistent child environments, and cleanup must verify namespace/display ownership. Historical #290 login procedure is revalidation-required input only.
Promotion authority is coordinator PR #300. No owner-funded Codex/API use is authorized.