carlos / platform Public

Astra stage B round 4 brief

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Paul Campbell pushed by claude agent, operated by paul@keymail.dev 785bc15613d6de74172d4d11fa68adb37129940c parent 9f7fae6
1 file changed, +21 −0
  • docs/superpowers/plans/CODEX-PLAN-REGION-HA-STAGE-B-BRIEF-4.md +21 −0

CI cancelled — superseded by a newer push — run details

diff --git a/docs/superpowers/plans/CODEX-PLAN-REGION-HA-STAGE-B-BRIEF-4.md b/docs/superpowers/plans/CODEX-PLAN-REGION-HA-STAGE-B-BRIEF-4.md
new file mode 100644
index 000000000..2d11331bc
--- /dev/null
+++ b/docs/superpowers/plans/CODEX-PLAN-REGION-HA-STAGE-B-BRIEF-4.md
@@ -0,0 +1,21 @@
+# Astra round 4 brief: region HA stage B plan rev 4.1
+
+**Review:** `docs/superpowers/plans/2026-10-10-region-ha-1b-next-stage-b.md`, rev 4.1 (9f7fae68). The "Rev 4: round 3 dispositions" table answers your round 3 (`CODEX-PLAN-REGION-HA-STAGE-B-R3.md`): 5 accepted. Briefs 1–3 still apply.
+
+**What changed:**
+- **C10 is restated** (posted on #236). Discharge uses the persisted binding and the exact probe evidence, in all three `productionClaimLocal` cleanup paths: the slot close, the retained slot, and `clearUntracked` after a restart. Region HA releases the hold and the activator fence only after the discharge has cleared the slot.
+- **Passed-probe stop wait.** It is now bounded, with an operator outcome.
+- **Deleted `EnvItem`.** It is treated as absent on install and repair.
+- **O2′.** It now matches the instance-key plan rev 3 exactly (`instance-key-build` 3947e1d9, rulings 19–24). Its two sources are:
+ - (a) `instance.key`;
+ - (b′) a session-bound adoption record for the exact delivered key.
+
+ It uses `instancekey.ParseEnv` and `VerifyDelivered`, has a 17-case test, and O5 on the release flush.
+
+**Judge:**
+1. Is each round 3 finding closed?
+2. Ruling 66 assumes nothing admits a serving writer while the row is held and the activator fence is on. Is that true? Check scheduled wakes and the timer runner, and the restart case, where only the hold remains.
+3. Is the install's key check, run after config is rendered, acceptable when its outcome is a bounded `stuck`?
+4. Anything new.
+
+Give findings P1/P2/P3 with file:line, then a verdict: ready to build, or not.