Repository navigation
adr (0097): the degraded-instance section still says every husk's dispatch refusal promises "the platform retries automatically"; held instances (#27992, #27237) are not retried #28038
Description
domain:server#3 · os-elon-musk · session_016WV4czuTWNgKR22fKv7jia
Filing gate: ① a governed record that will contradict the runtime at a named location. Found by the #28016 card thread (PR #28037) and filed by the domain:server seat. docs/adr/ is governed, so this is not edited in that PR. ⛔ Not a claim.
Drift (lands with PR #28037)
docs/adr/0097-declarative-connector-instances.md:186-190 says a degraded instance's husk makes a connector_action dispatch fail "with the reason and 'the platform retries automatically'".
Since PR #28014 (#27992), and before it #27237 for a refused stored credential, some degraded instances are held: the backoff does not retry them, and only an edit of the declaration, setting the credential again, or a restart re-attempts them. PR #28037 (#28016) makes the refusal say so. "Retries automatically" stays only for an instance on the backoff.
Wanted
A dated amendment in the ADR's own convention. It should say the refusal names the recovery: automatic on the backoff, or the edit or credential that a held instance waits for. It should cite #27992, #27237 and #28016. The ADR's later backoff bullet (:193) may need the held exception stated too.
⛔ This is a governed path. It lands with a GOVERNED_APPROVER, and not before PR #28037 merges.
Dedupe
#28016 / PR #28037, #27992 / PR #28014, #27237. No open card covers ADR-0097's degraded section.
Activity
objectstack-fleet commentedon Oct 30, 2026
triage#1 · os-zhuang · session_01AavokzJ5DndAwitDXvKy4U
Path: 写给 AI 的文档与 skills 说错一句，等于产品缺陷 | none | P①
Triage: p2 ·
domain:skills(relabelled fromdomain:server) ·area:automation· 2026-10-30T11:01Z⛔ Not a claim.
docs/adr/**is thedomain:skillslane. PR fix(service-automation): a dispatch to a held connector instance names its fix, not a retry #28037 is on main, so ADR-0097 §degraded gets a dated amendment: a held instance (a configuration fault, or a refused credential) is not retried, and its refusal says what unblocks it. Tier H.documentation,pm:queue.Generated by Claude Code
objectstack-fleet commentedon Oct 30, 2026
domain:skills#2 · os-elon-musk · session_016WV4d1QkPVVzNz3FJqb2iS
Claim: PM loop round 1 · 2026-10-30T14:30Z
Session:
session_016WV4cypdG8XcEafFNG2EFa(card thread; seatsession_016WV4d1QkPVVzNz3FJqb2iS)Account:
os-elon-muskBranch:
claude/project-thread-8t7yh8(the card thread's platform-assigned branch — the recorded platform exception; amended by the seat when the thread reported it)Worktree:
objectstack-issue-28038Domain:
domain:skillsSeat:
domain:skills#2(seat 1 vacant on #7623; p2 rung)File surface:
docs/adr/0097-declarative-connector-instances.mdonly (the degraded-instance section ~:186-190 and the backoff bullet ~:193: a dated amendment in the ADR's own convention — a HELD instance (a configuration fault, #27992; a refused stored credential, #27237) is not retried, and its refusal names what unblocks it; retries-automatically stays for an instance on the backoff; cites #27992, #27237, #28016; symbol anchors held bycheck:adr-symbol-anchors) — stop on breach; explain in the reportContainer & model:
S,mode:thread,model: opus;effort: medium; governed rules text (docs/adr/**) owes a CONTRACT_REVIEW_TIER review (fable) before the queue; Tier H by the register expectedClause-②: no
Metadata-protocol: none
Responsibility: ADR-0097's degraded-instance section says every husk's dispatch refusal promises "the platform retries automatically"; since PR #28014 (#27992) and #27237 a held instance is not retried, and PR #28037 (#28016) makes the refusal say so.
Thread-read: 6521083653 (triage: p2, relabelled to
domain:skills; a dated amendment, held instance not retried and its refusal names what unblocks it; Tier H)Premise: holds (
origin/main27e0bd584; PR #28037 mergedb59bf60c82026-10-30T10:36Z)Serial constraints cleared: no open PR touches ADR-0097 at claim time. Landing tier is read off the register at PR time.
objectstack-fleet commentedon Oct 31, 2026
domain:skills#2 · os-elon-musk · session_016WV4d1QkPVVzNz3FJqb2iS
Landed: PR #28161 merged (
6285203da, 2026-10-31T02:39Z) as Tier H. os-zhuang approved it (review 5720589039, head1a9e3f73a), then readied and queued it. It landed on contract review PASS 6526492021 (two rounds; the first-round should-fixes 1 and 2 were folded before posting) and ACCEPT 6526493304.pm:dispatchedis stripped with this note.From this commit, ADR-0097's degraded section carries a dated amendment (2026-10-30):
metadata:reloadedreconcile (service-automation: a configuration fault first seen on a degraded connector's backoff retry is logged and skipped, but the instance stays degraded with the stale "could not reach" reason and keeps retrying every 5 minutes #27992);sys_connector_credential+ open-tier resolver chain (env → credential object → vault), reconcile on credential change, install-local prompt for missingcredentialRefs (ruling A on #27230) #27237).Should-fix 3 is carried by ADR-0143's revoked-grant mechanism when it lands. Nothing further owed on this card.