From 069d05bc03d42e925b22c87a3e8b81634caf347a Mon Sep 17 00:00:00 2001 From: John McKean Date: Mon, 28 Sep 2026 08:04:10 +0800 Subject: [PATCH] Record the end of weekly GSC checks and today's clean-up lessons index-quality runbook: John decided on 2026-09-28 not to resume the weekly checks; #237 closed, recovery unproven. pm-delivery gotchas: exclude the default branch from merged-branch clean-ups, squash-merged branch deletion by PR head, git stash -u leftovers, and Sentry 11's permissive dataCollection defaults. Co-Authored-By: Claude Opus 5.5 --- docs/agents/pm-delivery.md | 9 +++++++++ docs/seo/index-quality-verification-2026-07-21.md | 4 ++-- 2 files changed, 11 insertions(+), 2 deletions(-) diff --git a/docs/agents/pm-delivery.md b/docs/agents/pm-delivery.md index 151ed422..9945c612 100644 --- a/docs/agents/pm-delivery.md +++ b/docs/agents/pm-delivery.md @@ -105,3 +105,12 @@ only that delta. doesn't prove the cause. - Setting a production environment variable is a production change: it goes live on the next deploy, whichever PR triggers it. +- In a branch clean-up, exclude the default branch by name. A stale local `main` counts as + "merged into origin/main", so a merged-branch list will include it. +- For a squash-merged branch, `git branch -d` reports "not fully merged". Delete it only when its + tip equals the head of the merged PR (`gh pr view --json headRefOid`). +- `git stash push -u` can leave some untracked files behind. Check `git status` before switching + branches, and back up anything that blocks the switch rather than forcing it. +- A major SDK upgrade can change privacy defaults. Sentry 11 replaced `sendDefaultPii` with a + `dataCollection` option whose defaults collect IPs, cookies, headers and bodies, so set it + explicitly (#290). diff --git a/docs/seo/index-quality-verification-2026-07-21.md b/docs/seo/index-quality-verification-2026-07-21.md index c303eb8a..303c15d2 100644 --- a/docs/seo/index-quality-verification-2026-07-21.md +++ b/docs/seo/index-quality-verification-2026-07-21.md @@ -30,6 +30,6 @@ The Page indexing report was last updated 10 July and remains the same dated bas ## Weekly observation runbook -Weekly checks ran from 29 Jul to 1 Sep 2026 and have since stopped. Resuming them is a decision tracked on issue #237. If resumed, record a dated snapshot with the Page indexing report's last-update date, indexed/not-indexed totals, the 99-URL Crawled cohort, the 18-URL Discovered cohort and the sitemap's last-read/discovered-page figures. Compare URLs against `docs/seo/gsc/2026-07-21/url-classification.csv`; investigate only changes in a current URL's response, canonical, robots directive or sitemap membership. +Weekly checks ran from 29 Jul to 1 Sep 2026. On 28 Sep 2026 John decided not to resume them, and #237 was closed as not planned; this runbook is historical. If they are ever restarted, record a dated snapshot with the Page indexing report's last-update date, indexed/not-indexed totals, the 99-URL Crawled cohort, the 18-URL Discovered cohort and the sitemap's last-read/discovered-page figures. Compare URLs against `docs/seo/gsc/2026-07-21/url-classification.csv`; investigate only changes in a current URL's response, canonical, robots directive or sitemap membership. -Do not claim recovery until Search Console's report date advances beyond this submission. At the post-recrawl checkpoint, close #237 only if sitemap indexation is at least 85% and Crawled — currently not indexed is below 50, or create follow-up issues that name and explain every remaining cohort. Do not manually request indexing in bulk; single-URL inspections are reserved for a diagnosed exception. +Do not claim recovery until Search Console's report date advances beyond this submission. The 85% sitemap-indexation and below-50 Crawled — currently not indexed thresholds were never re-measured, so recovery remains unproven. Do not manually request indexing in bulk; single-URL inspections are reserved for a diagnosed exception.