docs(adr): addendum to ADR-0008 after three days of production - #14
Merged
Conversation
The decision is unchanged and the record is not rewritten. What changed is that three statements in it are no longer true, and an ADR whose status block says "no production vintage has been captured yet" while the store holds 224,280 rows is the stale-status-line failure this workspace keeps paying for. So: a pointer under the Status block, and an addendum at the end carrying the corrections, the open questions now closed, and, separately, the ones closed as deliberate DEFERRALS rather than left dangling. Reopening a deferral is a new decision, not a resumption, which is the distinction that keeps an open-questions list from becoming a backlog nobody owns. Corrections: capture has been running daily on the producer since 2026-07-31; disaggregated/TFF canonicalisers shipped in cotdata #83; and zero revisions have been detected, so every claim here about revision behaviour is still untested against a real CFTC restatement, which is the honest status of the subsystem's central purpose. Also records that the deployment constraint the Consequences section predicted held up under contact, and that the first operational defect was in the alarm rather than the data (cotdata #90): an expectation written per week and evaluated per run. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
ADR-0008 was accepted 2026-07-30 with the work already done. Three days of production have made three statements in it false, most visibly a Status block reading "No production vintage has been captured yet" while the store holds 224,280 observation rows.
The decision itself is unchanged, so this does not rewrite the record. It adds a pointer under the Status block and an addendum at the end.
What the addendum carries
Corrections. Capture has run daily on the Windows producer since 2026-07-31, syncing outward. Disaggregated/TFF canonicalisers shipped in cotdata #83, closing the open question that says only Legacy is wired. And zero revisions have been detected so far, which is not a defect but does mean every claim here about revision behaviour is untested against a real CFTC restatement — the honest status of the subsystem's central purpose.
Open questions resolved, including one recorded outside this ADR: the
announcedrelease-date tier, whichcot_vintage.md§9 had as acceptance criterion 5 unmet on the grounds that reading a release date out of prose would be guessing. Sound reasoning, unchecked premise — CFTC publishes an exact table. Built in cotdata #91; 36,296 rows move off thederivedfallback.Open questions closed as deliberate DEFERRALS, which is the part that matters for hygiene: futures-and-options-combined, revision-driven cache invalidation, and retention are decided, not pending. Reopening one is a new decision rather than a resumption. An open-questions list that never distinguishes those becomes a backlog nobody owns.
Two things worth carrying. The deployment constraint the Consequences section predicted held up under contact, and now has a watchdog on the replica because the existing sync check watched a file the vintage task does not write. And the first operational defect was in the alarm rather than the data (cotdata #90): an expectation written per week and evaluated per run, with the design doc containing both halves of the contradiction in adjacent paragraphs.
🤖 Generated with Claude Code