The Gold Standard · Anthropology
Make the decision path visible
Quarantine is a protective pause, not a permanent home for every late-reconnecting device or service. The reviewer must decide which pending observations still describe real events, which writes depend on obsolete authority and what the consumer must prove before it can safely rejoin the appeal route.
The gold standard separates evidence from authority. It preserves useful local facts, rejects stale ownership and closure writes with reasons, then restores only the read and write capabilities supported by a current graph, a clean replay and an accountable release record.
Choose the route closest to your decision
Chapter contents
Chapters 1–3
Chapters 4–6
Chapters 7–9
Chapters 10–12
CHAPTER 1 OF 12 · CLASSIFY THE QUEUE
1. Define the quarantine decision
For each consumer, state why it was held, which graph version it presented, what pending changes exist and who can approve release. A quarantine without a decision owner can become a silent operational backlog.
Separate consumer release from individual item acceptance. A device may remain quarantined while a verified observation is admitted through another route.
CHAPTER 2 OF 12 · CLASSIFY THE QUEUE
2. Classify pending items by consequence
Group items as factual observations, contact details, delegation changes, ownership writes, acknowledgements or closures. Mark whether each item can affect who participates or what state becomes final.
This classification prevents a useful field note from being discarded merely because it travelled beside an obsolete authority change.
Evidence check
Check whether the scope, owner and decision consequence are explicit before continuing.
CHAPTER 3 OF 12 · CLASSIFY THE QUEUE
3. Establish current authority and evidence rules
Use the authoritative graph, effective-time record, delegation history and decision charter. Define what proof allows a fact to be retained and what proof would be needed to accept a consequential write.
A local timestamp, user confidence or successful login does not establish that the consumer held current decision authority.
| Item | Decision question |
|---|---|
| Observation | Can provenance be verified? |
| Contact change | Who confirms it? |
| Ownership write | Was authority current? |
| Closure | Could outcome change? |
CHAPTER 4 OF 12 · REJECT SAFELY
4. Reject stale writes with durable reasons
Return a structured rejection that names the obsolete version, affected field, current authority and next review route. Preserve the attempted write and its source sequence.
Silent dropping hides both recurrence and legitimate evidence. A durable rejection lets the owner repair the consumer and lets auditors reconstruct the attempted change.
CHAPTER 5 OF 12 · REJECT SAFELY
5. Prevent rejected changes from reappearing
Clear or mark the local queue, update cached records and test retry behaviour. Block connectors that repeatedly resubmit the same stale operation under a new timestamp.
A central rejection is incomplete if the consumer treats it as a transient network error and keeps trying.
Evidence check
Ask a second reviewer to reconstruct the example from the preserved evidence and stated assumptions.
CHAPTER 6 OF 12 · REJECT SAFELY
6. Worked example: a contractor app carries facts and stale ownership
A contractor app reconnects with two photographs of a community meeting, a corrected phone number and a closure assigned to a delegate whose authority ended during the offline period. All four items share the same old graph version.
The reviewer retains the photographs after provenance checks, verifies the phone number through the current owner and rejects the closure. The app is updated before release. Changed-context check: if the ended delegate recorded a factual acknowledgement while still authorised, that receipt can remain evidence even though the later closure cannot.
| Field | Purpose |
|---|---|
| Presented version | Shows stale state |
| Rejected write | Defines boundary |
| Current authority | Explains correction |
| Next route | Supports resolution |
CHAPTER 7 OF 12 · PROVE REINTEGRATION
7. Build a clean replay against current state
Load a current snapshot into an isolated environment and replay retained items using their original order. Confirm that obsolete writes fail, accepted facts land in the correct fields and no duplicate action occurs.
Replay should test both success and refusal. A consumer that can only demonstrate happy-path syncing has not proved safe authority behaviour.
CHAPTER 8 OF 12 · PROVE REINTEGRATION
8. Release the minimum capability first
Restore read access, then bounded factual writes, then any consequential write path only if it is necessary and independently verified. Record who approved each stage and when it expires for review.
Minimum capability reduces the impact of an undiscovered local cache or retry loop while useful work resumes.
Evidence check
Verify that the control still reaches the intended decision under a changed but plausible context.
CHAPTER 9 OF 12 · PROVE REINTEGRATION
9. Verify safe reintegration in a changed context
Test another graph change, an expired credential, a second offline interval and a rejected item followed by a valid one. Confirm the consumer refreshes authority rather than copying the old state forward.
The test should include the people who resolve the queue so error messages lead to a practical next step.
| Stage | Proof |
|---|---|
| Read | Current graph loaded |
| Factual write | Clean replay |
| Consequential write | Independent test |
| Full release | Owner approval |
CHAPTER 10 OF 12 · KEEP RE-ENTRY SAFE
10. Monitor post-release behaviour
Watch version mismatches, repeated rejections, unexpected ownership fields, duplicate receipts and writes from unregistered identities. Set a threshold for automatic return to quarantine.
Use the longest credible offline cycle when judging stability; the absence of immediate errors does not prove that delayed queues are gone.
CHAPTER 11 OF 12 · KEEP RE-ENTRY SAFE
11. State limits and misconceptions
Safe release cannot prove that every local export, printed list or informal delegation was found. Staged access can also slow urgent field work when review capacity is weak.
Record untested paths, emergency arrangements and any capability kept disabled. Reintegration is a bounded evidence claim.
Evidence check
Confirm that versions, notices and unresolved limits remain visible in the final record.
CHAPTER 12 OF 12 · KEEP RE-ENTRY SAFE
12. Audit acceptance, rejection and release
Track each pending item, classification, evidence, decision, replay result, capability stage, release owner and later recurrence. Compare the consumer inventory with actual writers.
The gold standard ends quarantine with a reasoned state: useful evidence is preserved, stale authority is refused and the returning consumer earns only the capability it can safely exercise.
| Signal | Warning |
|---|---|
| Repeated rejection | Queue not repaired |
| Duplicate receipt | Replay fault |
| Old owner write | Authority leak |
| Unknown identity | Inventory gap |
Frequently asked questions
Must every quarantined item be rejected?
No. Preserve verified factual evidence separately from writes that require current decision authority.
Why record a rejection instead of deleting the item?
The record explains what failed, prevents silent retry and preserves evidence for repair and audit.
Can a consumer return in read-only mode?
Yes. Staged release is often the safest way to restore useful access while consequential writes remain disabled.
When is reintegration complete?
After current-state replay, changed-context testing, bounded capability release and a monitoring period with no material recurrence.
Continue through verified routes
Core routes
Anthropology subject root · Gold Standard of Education · Gold Standard directory · How X Works hub
Specialist routes
- Contact-graph re-synchronisation drift, late-reconnecting consumers and recurrence containment
- Contact-graph rollback reconciliation, split-brain directories and authoritative re-synchronisation
- Contact-graph drift repair rollback, incorrect synchronisation and restoring the last known-good route
- Contact-graph version drift, stale alternates and recurrence prevention