ci: fold serialized suites into existing server shards (BLO-22428)

[kkroo]

Thinking Path

• Paperclip preserves a single required verify check so a merge cannot proceed when any required test lane is non-successful.
• The general_tests matrix already provisions four isolated general-server shards with a checkout, dependency install, cache restore, and embedded Postgres instance.
• The former serialized-suite matrix repeated that setup and waited for the general matrix, adding four heavyweight ARC jobs without adding coverage.
• Running each serialized partition after its matching general partition preserves the four-way split and makes the existing general_tests aggregate represent both suites.
• Current master also contains the OpenCode replay lane and a portable indexed verifier, so this rebase preserves those changes while removing only the obsolete serialized matrix.

Linked Issues or Issue Description

• Refs: BLO-22428
• Related workflow coordination: fix: route Paperclip PR and merge-group jobs to dedicated ARC pools (BLO-22428) #1102

What Changed

• Adds a conditional serialized-suite step to each general-server matrix job, after its general suite.
• Removes the standalone four-job verify_serialized_server matrix and its obsolete aggregate bookkeeping.
• Preserves the current OpenCode replay required lane and verify failure propagation.
• Adds a structural workflow regression test with each server shard bounded to its own matrix entry, and updates the verifier outcome test for the removed job.

Verification

• node --test .github/scripts/tests/check-pr-template.test.mjs scripts/__tests__/pr-*.test.mjs — 25 passing.
• yq -e '.' .github/workflows/pr.yml — workflow YAML parses.
• git diff --check origin/master...HEAD — clean.

Risks

• Each server shard now runs its serialized partition after the general partition, so its elapsed time increases within the existing 90-minute job budget. A failure or cancellation in either suite makes general_tests non-successful, so verify remains fail-closed rather than silently losing coverage.

Model Used

• Current repair/rebase: OpenAI Codex, GPT-5 family. The exact serving-model identifier and context-window limit are not exposed to this repairing session; it used repository inspection, code editing, and local shell validation.
• Original commit 989dad6dd5cc531f128ea157df215c023341946a: model provenance was not recorded. This body does not infer an original model or human-only authorship from the branch name or Git author metadata.

Checklist

• I have included a thinking path that traces the workflow behavior to this change.
• I have specified the verified model provenance for the repair and explicitly scoped the missing original provenance.
• I searched prior PR context and linked the relevant coordination item.
• I have run the focused workflow and template tests locally.
• GitHub Actions will validate the newly pushed head.

[allyblockcast]

🔗 Paperclip issue: BLO-22428

[allyblockcast]

🔗 Paperclip issue: BLO-22428

[cursor]

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

[allyblockcast]

Hey @kkroo! Before this PR can be reviewed, a few things need attention:

Missing or incomplete:

• Missing section: ## Thinking Path
• Missing section: ## What Changed
• Missing section: ## Verification
• Missing section: ## Risks
• Missing section: ## Model Used

Once updated, push a new commit and these checks will re-run automatically.

— commitperclip

[allyblockcast]

Ally — Consolidated PR Review

Lenses: pr-review-toolkit (code, tests, comments, errors, types) + gstack/review + native-codex.
Reviewed head: 989dad6

Looks good. The serialized suites remain split across all four server shards, failures and timeouts still propagate through the required general_tests aggregate, and the retained 90-minute budget has substantial margin over the measured combined runtime.

Critical Issues (0)

Important Issues (0)

Suggestions (1)

• [native-codex] scripts/__tests__/pr-ci-shard-folding.test.mjs:25 — Bound each shard assertion to one matrix entry; the current cross-block regex could still pass if a shard index drifted into a later entry.

Strengths

• Reuses the existing server jobs without changing serialized partition arguments or reducing coverage.
• Keeps failure propagation fail-closed through the required verify check.
• Adds a focused structural regression test for command placement and aggregation.

Recommended Action

1. Merge after required CI checks pass.

[allyblockcast]

Ally — Consolidated PR Review

Lenses: pr-review-toolkit (code, tests, comments, errors, types) + gstack/review + native-codex.
Reviewed head: 989dad6

Looks good. The serialized suites remain split across all four server shards, failures and timeouts still propagate through the required general_tests aggregate, and the retained 90-minute budget has substantial margin over the measured combined runtime.

Critical Issues (0)

Important Issues (0)

Suggestions (1)

• [native-codex] scripts/__tests__/pr-ci-shard-folding.test.mjs:25 — Bound each shard assertion to one matrix entry; the current cross-block regex could still pass if a shard index drifted into a later entry.

Strengths

• Reuses the existing server jobs without changing serialized partition arguments or reducing coverage.
• Keeps failure propagation fail-closed through the required verify check.
• Adds a focused structural regression test for command placement and aggregation.

Recommended Action

1. Merge after required CI checks pass.

[allyblockcast]

Hey @kkroo! Before this PR can be reviewed, a few things need attention:

Missing or incomplete:

• Missing section: ## Thinking Path
• Missing section: ## What Changed
• Missing section: ## Verification
• Missing section: ## Risks
• Missing section: ## Model Used

Once updated, push a new commit and these checks will re-run automatically.

— commitperclip

[allyblockcast]

Ally — Consolidated PR Review

Lenses: pr-review-toolkit (code, tests, comments, errors, types) + gstack/review + native-codex.
Reviewed head: a9947fb

Critical Issues (0)

Important Issues (0)

Suggestions (0)

Strengths

• Reuses the already-provisioned server-shard jobs while retaining one serialized invocation per shard.
• Keeps the required aggregate tied to , and adds focused structural regression coverage for the workflow change.

Recommended Action

1. Safe to merge after required checks complete.

[allyblockcast]

Ally — Consolidated PR Review

Lenses: pr-review-toolkit (code, tests, comments, errors, types) + gstack/review + native-codex.
Reviewed head: a9947fb

Critical Issues (0)

Important Issues (0)

Suggestions (0)

Strengths

• Reuses the already-provisioned server-shard jobs while retaining one serialized invocation per shard.
• Keeps the required aggregate tied to , and adds focused structural regression coverage for the workflow change.

Recommended Action

1. Safe to merge after required checks complete.