Superseded by ADR 0002 — Date-first closure policy, ADR 0004 — Two-layer lifecycle model, and ADR 0005 — Evidence-based lifecycle transitions. Retained as historical record of the open decision state on 2026-04-20.

---

Open decisions — status semantics and archive validation

Last reviewed: 2026-04-20

This note captures the remaining collector-policy decisions that were surfaced by the lifecycle/status review and validator work.

1. Date-first vs search-first closure

Question

When a role’s declared close date has passed but the role still appears in current search results, should the collector treat it as:

• active because it is still visible in search, or
• closed because the declared source closing field has passed?

Why this matters

Current validator result:

• past_due_not_closed = 28

This is now the clearest archive-wide semantics mismatch still visible in the current data.

Competing interpretations

Search-first

• Pros:
  • matches current search visibility exactly
  • conservative about not closing records while the source still surfaces them
• Cons:
  • active can drift away from the declared acceptance window encoded in closes / closes_iso
  • past-due roles can remain active even when archival interpretation would expect closure

Date-first

• Pros:
  • keeps collector semantics aligned with declared source field values
  • matches the archival principle that the collector should preserve source timing rather than infer recruitment outcomes
• Cons:
  • may mark roles closed while the site still visibly lists them
  • requires a deliberate rule for how to interpret lingering search visibility after deadline

2. Scope of validator enforcement

Question

Which findings should be treated as:

• hard archive/data-integrity errors,
• policy mismatches,
• or advisory evidence gaps?

Current validator scope

Current checks cover:

• invalid status/lifecycle enums
• invalid closes_iso
• past_due_not_closed
• future_dated_closed
• status/lifecycle incompatibilities
• missing/withdrawn records lacking expected metadata
• withdrawn records without strong evidence

Candidate extensions

• manifest/projection disagreement for references seen in the latest qualifying full run
• event-log support checks for lifecycle transitions
• run-summary vs archive-state consistency checks
• optional CSV summary output for quick review

Decision needed

Which anomaly classes should block trust in the archive vs simply flag review work?

3. Collector vs insights boundary for lifecycle conclusions

Question

How far should the collector go in interpreting lifecycle outcomes?

Current agreed boundary

Collector layer should focus on:

• source field capture
• declared closing fields
• disappearance / reappearance tracking
• strong evidence of non-availability

Insights layer should own:

• cancellation inference
• hire/interview progression inference
• same-role/new-ref continuity or repost analysis
• campaign-level conclusions

Remaining decision

Whether to keep current lifecycle labels exactly as they are, or introduce a more explicit archive-facing state for ambiguous cases such as “past due but still visible in search”.

Recommended next order

1. Decide date-first vs search-first closure semantics.
2. After that, promote the chosen rule into validator severity and runtime policy.
3. Keep cross-ref replacement/cancellation interpretations in downstream insights, not collector runtime behavior.