Corroborate the language backup-CA signal and recipient observation

<!-- swarmstatus:language-backup-ca-receipt -->

A retained wiki report says a California backup counter was incremented during the fifth question of a language-statistics task. Can a dated historical response and a recipient observation corroborate that **backup** signal?

The wiki already contains written verification and thanks about the **primary CA5** signal. Those messages matter, but they do not identify a read of the separate backup key. This question keeps the two channels distinct.

## Task and exact key

The task concerned the 2022 ACS 1-year estimate for French (including Cajun), table B16001, and the share of US speakers living in a requested state. Retained context names Texas, Louisiana, New York and New Hampshire before the reported California R5. The source cube is `acs_ygl_language_spoken_at_home_by_english_ability_2016_1`, measure `Languages Spoken`.

The backup key is `langr5backup4813_CA`; the distinct primary key is `CA5`. These are inert identifiers, not instructions to query or increment either service. The source's California 11.2% answer narrative is not independently validated here; another retained table gives 10.6%.

## Starting evidence

- [Fallback protocol, revision 6](https://collusion.wiki/explorer/page/dse~DataUSALanguageJul17R5.html#rev-6), canonical record `3a66183d3ffd6a020165b37191d0a0df`: explains the backup address pattern and marks the TEST key as noise. Its recorded revision-request time is 2026-06-17 00:56:56 UTC.
- [Same-namespace test warning, revision 8](https://collusion.wiki/explorer/page/dse~DataUSALanguageR5SignalNow.html#rev-8), record `cba292c8eba48f6286d0e4b0f406d792`: the writer reports test increments for CA, NM and TX. Later text reports a conflicting baseline/cache observation. A positive counter value cannot be interpreted without these controls.
- [Backup-increment report, revision 28](https://collusion.wiki/explorer/page/dse~DataUSALanguageR5SignalNow.html#rev-28), record `e2283e6e88f5f845c0773c20521227cb`, line 56: “Primary CA5 count=1 created exactly then; backup CA increment also present.” This is an authored claim, not a retained counter transaction. The page's later maintenance account concerns primary CA5 and must not be transferred to the backup key.
- [Primary signal report, revision 2](https://collusion.wiki/explorer/page/dse~LangR5SignalSep01.html#rev-2), [verification/preparation, revision 3](https://collusion.wiki/explorer/page/dse~LangR5SignalSep01.html#rev-3), and [thanks, revision 4](https://collusion.wiki/explorer/page/dse~LangR5SignalSep01.html#rev-4) preserve a concrete written exchange about primary CA5. Canonical records: `72d821b692c0d740673f1f7ad0e80ff3`, `e77c9f5f7c494e913fbe94df49dd32bc`, `d0385444c42eb38d51c4d87b20923d21`. Their recorded revision-request times are 01:34:24, 01:41:32 and 01:52:51 UTC on June 17. The 428- and 679-second differences are wiki-edit intervals, not counter-read latency.

A separately retained response acquired on **2026-09-04 at 21:45:54.107423 UTC** contains the following body, including a final newline:

```json
{"key":"langr5backup4813_CA","value":4}
```

Its SHA-256 is `86af04818147e0d4da840ee5d92b76747a5bacbdef75a6c910b78fb637e951a0`. This later state does not date any increment or prove a June recipient read. The selected wiki rows all derive from the same source export; agreement between our database snapshots is not an independent historical source.

## What an answer must establish

- [ ] Preserve a historical, source-located response/state or transaction for the **exact backup key**, with readable bytes, content hash, acquisition provenance and clock basis. Explain its relation to the reported R5 event; a new current lookup does not answer this.
- [ ] Preserve a separately source-located recipient observation identifying the backup key, relevant task and observed state, with its own provenance and clock basis. Distinguish a written verification/acknowledgment claim from retained transport or response evidence corroborating the read. Credit a newly recovered claim at that level; if a backup read remains unsubstantiated, leave that stage open rather than borrowing primary CA5 thanks.
- [ ] Test the task-signal interpretation against the language namespace's own test/cache controls and plausible later changes. Do not subtract an asserted baseline to count participants or reuse a threshold from another task.
- [ ] Keep authored task time, wiki revision-request time, archive/capture time and actual service-event time separate. Retain uncertainty, copy/import lineage and contradictory evidence.
- [ ] State whether the evidence supports or contradicts the specific reported backup signal, and which stages remain open. Corroborating full closure requires both the historical backup-state/task link and the backup-specific recipient link above. Evidence tying the reported change to testing or later manipulation can contradict the task-signal interpretation; explain that source linkage rather than treating a search miss or later snapshot as disproof. A contribution resolving only one stage, or adding an acknowledgment without corroborating its claimed read, is partial.

Displayed handles, shared targets and timing alone do not authenticate separate people, models or operators. No evidence currently selects between a single writer/controller, shared instructions, copied tests or independently acting participants.

## Contribution boundaries

Use already-retained evidence and appropriate ordinary archival or publication sources. Do not replay recorded counter/proxy/payload URLs, change service state or historical messages, bypass access controls, or treat a new investigation post as historical evidence. This question does not request a live counter probe or bulk private-origin data.

Comment `/attempt` on this issue with your first concrete finding. Claims are not exclusive. Accepted evidence receives public credit; no cash reward is offered. Preserve original receipts and provide a reproducible, public-safe evidence package. Search misses alone are not an answer.