Plan the encoded-ref probe - #249
Merged
Merged
Conversation
Every layer is tested and the golden vectors match across repos, but the full path has never carried a real respondent. This plans the legs that would prove it, cheapest first, and records what is already settled so none of it is re-measured. Two findings that shape it, both verified rather than assumed: ad_attributions holds zero rows. Rows are written only on ad creates, so that is consistent with a healthy write path with nothing to do -- and equally with record_ad_attribution never firing. Leg 0 tells those apart, because every later leg assumes the first. vlab staging cannot host the end-to-end test the handover assumed: it pins adopt v0.0.106 and has no migrations block, so the hook would template an empty image. The probe runs against production with a PAUSED ad instead. Also records the observability gap retiring the organic counter opened: a correctly-configured encoded study whose refs stop round-tripping now emits no signal at all. That is a decision to make, not a task to schedule. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
✅ Deploy Preview for vlab-dashboard canceled.
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Adds
planning/encoded-ref-probe-plan.md. Follows the idiom ofplanning/ctwa-probe-runbook.mdand picks up whereplanning/encoded-ref-handover.md§6c leaves off.Four legs, cheapest first, plus one descoped and said out loud rather than
implied. §1 records what is already settled with citations, so none of it gets
re-measured.
Two findings that shape it, both verified against vprod rather than assumed:
ad_attributionsholds zero rows. Rows are written only on ad creates,so that is consistent with a healthy write path that has had nothing to do —
and equally with
record_ad_attributionnever firing. Nothing distinguishesthem from outside, and every later leg assumes the first. That is leg 0.
adopt v0.0.106against production'sv0.1.79, andtoixo-staging.yamlhasno
migrations:block, so the pre-upgrade hook would template an empty image.The plan runs against production with a
PAUSEDad instead — the posture the2026-08-17 CTWA probe used successfully.
It also records the observability gap that retiring the organic counter opened:
a correctly-configured encoded study whose refs stop round-tripping now emits no
signal at all, and
thins_its_ref_without_reading_the_mappingdoes not catch itbecause that fires only when a study declares no lookup conf. Written up as a
decision with two options rather than a task, since accepting it is defensible
while the feature is one study wide.
Docs only — no code, no behaviour change.
🤖 Generated with Claude Code