Repository navigation
Refactor workflow email tests into readable shared scenarios #319
Description
Question
How can the workflow email tests read as clear behavioral scenarios, with shared helpers, objects and classes removing repeated wiring while preserving their coverage and agreed public seams?
Scope
Refactor the email test code from the existing stack, based on refactor/workflows-email-pacing (PostHog#112897). Review the S1 service suites, S2 worker round-trip and queue tests, S3 pacing policy tests, and their test helpers. Extract shared fixtures, service construction, invocation builders and resource lifecycle where they earn their keep. Keep scenario-specific arrangements and assertions visible. Preserve all existing cases, coverage, real SDK requests and queue round-trip behavior. Production code stays unchanged.
Delivery
One reviewed draft PR from the Silthus fork to PostHog/posthog, targeting master and explicitly stacked on PostHog#112897. Do not rewrite the earlier stack or touch PostHog#112295 or PostHog#112071. Follow the existing agreed seams S1–S3, run before/after affected suites and the strict local gate, and include proof and a merge-risk grade in the PR template. Link the PR to the conductor thread.
Origin
Continuation of the email send path and SES testing map. Requested test maintainability cleanup; behavior and production delivery are unchanged.
Activity
Silthus commentedon Oct 7, 2026
Runtime: T3 Code delegated Codex child (GPT-6.1 Sol)
Task: node:delegated-task:command%3Amcp%3A2ff98002-8d9c-4a45-9e98-24c03c73f81d%3Adelegate-task%3Awf291-test-cleanup-319-r1
Branch:
chore/workflows-email-test-fixtures, based onrefactor/workflows-email-pacingat7938ae5dfd556.Cap: 90 minutes, enforced by a one-shot scheduler wake that cancels unfinished work and deletes its own schedule.
This ticket preserves the agreed S1–S3 seams and the existing behavior cases. It adds a test-only draft layer on top of PostHog#112897. The outcome refactor waits for this cleanup as well as the already recorded sandbox gate.
Cross-family review availability will be checked live by the worker. The shared availability report currently reports Claude exhausted; an unavailable review is recorded as incomplete, never claimed as run.
Silthus commentedon Oct 7, 2026
The implementation worker is terminal and its deadline wake has been removed. The branch is clean at
ed983c015d484b306386021a543807c403ed67c9; the linked draft is PostHog/posthog#113056, stacked on PostHog#112897.Implementation and local validation are available in the pinned proof bundle: shared typed email fixtures, SES request objects and focused queue/retry helpers; eight test/helper files only. Before and after runs passed 539 parallel tests and 83 serial tests with identical names, zero skips and 684 retained normalized assertions. Lint, typecheck, formatting and strict preflight passed.
The ticket remains open because the required cross-family review has not run. The native Opus review reported a rate limit; the prescribed CLI returned HTTP 429. CodeRabbit was signed out and skipped. No review findings exist to disposition. The shared availability check still reports Claude exhausted, with no reset time supplied. This is a provider blocker, not a pending human decision.
Resume by reviewing only this increment against parent
7938ae5dfd556818b61a65517d40410eac511411with read-only Opus. Validate each finding, fix what holds, record reasons for rejected findings, rerun affected checks after fixes and push normally. The ticket can close only after that review and the green local gate. Earlier PRs and the existing R3 sandbox gate remain intact.Silthus commentedon Oct 7, 2026
Runtime: existing T3 Code delegated Codex child, implementation follow-up (GPT-6.1 Sol).
Task: node:delegated-task:command%3Amcp%3A2ff98002-8d9c-4a45-9e98-24c03c73f81d%3Adelegate-task%3Awf291-test-cleanup-319-r1
Follow-up run: ordinal 2.
Cap: 45 minutes, with a one-shot deadline wake that deletes itself and cancels unfinished work.
The requested follow-up audits useless comments in both code and PR descriptions. PRs remain open. The conductor shortened all nine descriptions, retained evidence and review findings, and removed repeated process narration and template boilerplate. Risk/count/release markers remain functional.
The existing worker owns code cleanup in the current top layer. Scope now includes comment-only edits to production email files as well as test files. It must preserve executable tokens, useful external-contract and timing explanations, and all earlier branch histories. Final counts and the current PR description will be reconciled after it pushes. The required Opus review remains blocked by provider quota.
Silthus commentedon Oct 7, 2026
The comment-cleanup follow-up exceeded its cap and was interrupted. Its commit and push completed before the deadline:
e0e76088af5ab95991cc9582f0400ce012c7b56eis both the clean local head and the head of draft PostHog/posthog#113056. No child turns remain active, and the deadline schedule was deleted.The committed-head proof verifies identical executable tokens, syntax trees and assertion chains across all 22 audited files. Formatting and strict preflight passed. The conductor corrected the earlier formatting check's relative paths and verified its successful result. Runtime cases were not repeated for comment-only edits.
All nine PR descriptions were shortened, preserving evidence, risk/count markers and review findings. The current layer now spans six production files with comment-only edits and thirteen test/helper files. The PR description reflects its final head and counts. Earlier branches and PR states remain intact.
This ticket remains open solely for the required Opus review of the complete layer against its R2 parent. The provider remains exhausted, and no cross-family review has run. CodeRabbit was signed out and skipped. This is an external review blocker, not a question awaiting the human. No PR was closed or merged.