Records the live reproduction of #698's route 2, which #708 split out because it could not be
performed in the originating session.
Result: route 2 is REFUSED on the live instance after PR #705. Application code was pushed onto
Renovate PR #761's branch; the classifier logged
Changed files (2, complete=yes):
Directory.Packages.props
ErsatzTV.Core/Probe708.cs
Decision: state=pending — authored by the 'renovate' bot account, but changes files outside the dependency-manifest set — a bot ACCOUNT does not attribute the CODE at this head (the account is the PR's immutable creator; the head is not), so this needs a real verdict
Posted review-verdict/h10=pending on 830a407.
and the classify step's start marker was present, so this is a real classification rather than the #751 silent no-op. #698 does not need reopening.
PR #761 was guarded with a WIP: prefix (draft: true / mergeable: false confirmed) before anything
was pushed, then force-restored to its recorded sha 8798a1d2 and verified byte-identical across
title, head sha, base ref, draft, mergeable, state, labels, body hash, author and milestone.
Two corrections this also lands, because both would mislead the next reader:
The refusal reason is only in the job log.#708's step 5 expected the status description to say
the PR changes files outside the manifest set. It does not — reason is only echoed, and every
non-exempt head that is not carrying the raced-repair sentinel gets the generic Awaiting review verdict for <sha7>.
A live instance of #763, evidence posted there. Re-running the gate on #761's original head
tripped #751's page-2 "assume raced" rule and repaired a correct exemption to pending with a false ::error:: claiming a human verdict was overwritten; none existed.
Docs-only: one decision record, ci.exemption-provenance.
Verification: decisions_validate OK; catalog regenerated with no diff. Two rounds of cold,
worktree-isolated review — round 1 BLOCKED on an ambiguous head reference and an overstated
"every non-exempt head" invariant, both fixed; round 2 CLEAN, with the new causal claim
independently verified against the workflow source.
Records the live reproduction of #698's route 2, which #708 split out because it could not be
performed in the originating session.
**Result: route 2 is REFUSED on the live instance after PR #705.** Application code was pushed onto
Renovate PR #761's branch; the classifier logged
```
Changed files (2, complete=yes):
Directory.Packages.props
ErsatzTV.Core/Probe708.cs
Decision: state=pending — authored by the 'renovate' bot account, but changes files outside the dependency-manifest set — a bot ACCOUNT does not attribute the CODE at this head (the account is the PR's immutable creator; the head is not), so this needs a real verdict
Posted review-verdict/h10=pending on 830a407.
```
and the classify step's start marker was present, so this is a real classification rather than the
#751 silent no-op. #698 does **not** need reopening.
PR #761 was guarded with a `WIP:` prefix (`draft: true` / `mergeable: false` confirmed) before anything
was pushed, then force-restored to its recorded sha `8798a1d2` and verified byte-identical across
title, head sha, base ref, draft, mergeable, state, labels, body hash, author and milestone.
Two corrections this also lands, because both would mislead the next reader:
- **The refusal reason is only in the job log.** #708's step 5 expected the status *description* to say
the PR changes files outside the manifest set. It does not — `reason` is only echoed, and every
non-exempt head that is not carrying the raced-repair sentinel gets the generic
`Awaiting review verdict for <sha7>`.
- **A live instance of #763**, evidence posted there. Re-running the gate on #761's *original* head
tripped #751's page-2 "assume raced" rule and repaired a correct exemption to `pending` with a false
`::error::` claiming a human verdict was overwritten; none existed.
Docs-only: one decision record, `ci.exemption-provenance`.
Verification: `decisions_validate` OK; catalog regenerated with no diff. Two rounds of cold,
worktree-isolated review — round 1 BLOCKED on an ambiguous head reference and an overstated
"every non-exempt head" invariant, both fixed; round 2 CLEAN, with the new causal claim
independently verified against the workflow source.
fixes #708
Pushed application code onto Renovate PR #761's branch and captured the
review-verdict/h10 the gate posted: pending, with the classifier logging
"authored by the 'renovate' bot account, but changes files outside the
dependency-manifest set". Route 2 of #698 is refused on the live instance
after PR #705; identity alone no longer exempts a head.
Also records two things the issue got wrong or did not anticipate: the
refusal reason surfaces only in the job log, never in the status
description, and re-running the gate on the PR's ORIGINAL head tripped
#751's page-2 "assume raced" rule, which is the live instance of #763
that issue predicts.
Cold review found two defects in the first draft, both fixed here: the
repair paragraph said "that head", which reads as the probe head when the
repair in fact fired on the original head from the title edit five minutes
earlier; and "every non-exempt head gets the generic description" was false,
since the raced-repair sentinel is also a non-exempt description.
PR #761 was restored to its recorded sha and verified byte-identical.
fixes#708
Decisions-Edit: yes
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Two rounds of cold worktree-isolated review; round 1 BLOCKED on an ambiguous head reference and an overstated invariant, both fixed; round 2 CLEAN with the new causal claim independently verified against review-verdict.yml.
Review-verdict: MERGEABLE @ 8b8bb1c
Two rounds of cold worktree-isolated review; round 1 BLOCKED on an ambiguous head reference and an overstated invariant, both fixed; round 2 CLEAN with the new causal claim independently verified against review-verdict.yml.
timothy
merged commit e8e17f3844 into main2026-08-26 11:11:24 +02:00
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Records the live reproduction of #698's route 2, which #708 split out because it could not be
performed in the originating session.
Result: route 2 is REFUSED on the live instance after PR #705. Application code was pushed onto
Renovate PR #761's branch; the classifier logged
and the classify step's start marker was present, so this is a real classification rather than the
#751 silent no-op. #698 does not need reopening.
PR #761 was guarded with a
WIP:prefix (draft: true/mergeable: falseconfirmed) before anythingwas pushed, then force-restored to its recorded sha
8798a1d2and verified byte-identical acrosstitle, head sha, base ref, draft, mergeable, state, labels, body hash, author and milestone.
Two corrections this also lands, because both would mislead the next reader:
the PR changes files outside the manifest set. It does not —
reasonis only echoed, and everynon-exempt head that is not carrying the raced-repair sentinel gets the generic
Awaiting review verdict for <sha7>.tripped #751's page-2 "assume raced" rule and repaired a correct exemption to
pendingwith a false::error::claiming a human verdict was overwritten; none existed.Docs-only: one decision record,
ci.exemption-provenance.Verification:
decisions_validateOK; catalog regenerated with no diff. Two rounds of cold,worktree-isolated review — round 1 BLOCKED on an ambiguous head reference and an overstated
"every non-exempt head" invariant, both fixed; round 2 CLEAN, with the new causal claim
independently verified against the workflow source.
fixes #708
Review-verdict: MERGEABLE @
8b8bb1cTwo rounds of cold worktree-isolated review; round 1 BLOCKED on an ambiguous head reference and an overstated invariant, both fixed; round 2 CLEAN with the new causal claim independently verified against review-verdict.yml.