Preregister LEARN-0017: Zulip target-lookup/anchor continuation guard question

Bounded follow-up to ZN07 (ZulipNavigationSourceStudy.md), same static
pinned-revision protocol as LEARN-0014. No inspection performed yet.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

Assistant: claude-code
Assistant-Model: sonnet
Assistant-Process: 290719@bnt-lap001
Assistant-Session: 55cbe504-1bfe-4932-b761-29016523c87b

research/evolution/cycles/2026-09-27-zulip-target-lookup.md

@@ -0,0 +1,46 @@
+# LEARN-0017 — Target-message lookup and anchor-navigation continuation ownership
+
+Status: preregistered before inspection. Codex author-model review. 2026-09-27
+Europe/Berlin. Stream B, comparison group B-independent-navigation-source-v1
+(second cycle in this group; shares LEARN-0014's static-source-only protocol,
+evidence tier and effort accounting).
+
+- Date and owner: 2026-09-27, Codex.
+- Stream: B. Comparison group: B-independent-navigation-source-v1 (LEARN-0014's
+  static, read-only, pinned-revision source-inspection protocol).
+- Question and affected CAND identities: does independently developed Zulip code
+  guard the target-message lookup (`channel.get` to `/json/messages/:id` inside
+  `message_view.show`) and the anchor-navigation fetch continuation
+  (`navigate_to_anchor_message` / `message_fetch.load_messages_around_anchor`)
+  against a stale navigation that occurred while the request was in flight, the
+  way `message_view.ts`'s main narrow-fetch continuation does (ZN02)? Affects
+  CAND-0009 (URL-addressable State) and CAND-0010 (Progressive Enhancement).
+- Previous belief: ZN07 (ZulipNavigationSourceStudy.md) located both callbacks
+  and noted they have "no visible captured-current-list guard like ZN02," but
+  did not trace their actual callers or check test coverage.
+- Observation that would contradict the belief: either callback checking list/
+  narrow identity before acting, or an existing upstream test exercising a
+  stale-navigation race for either path.
+- Evidence/contexts to inspect: `web/src/message_view.ts` (target lookup ~L640-700,
+  `navigate_to_anchor_message` ~L998-1074), `web/src/message_fetch.ts`
+  (`load_messages_around_anchor`, `process_result`, `get_messages_success`,
+  `load_messages` error-path guard), `web/tests/message_fetch.test.cjs`,
+  `web/tests/message_view.test.cjs`. Same pinned revision as LEARN-0014/0016
+  (`5b06fa3d34f28fa01f459fbf3ba053e3c0858e9c`), reused deliberately. Out of
+  scope: deferred editing focus, search callback wiring, any application/test
+  execution, any defect or reproduction claim.
+- Effort budget and recording method: single-agent session, static reading only;
+  active research minutes recorded by wall-clock start/end, no execution
+  subprocess beyond repository checks (`make evolution-check evolution-test
+  schema-test`).
+- Review provenance and acceptance criteria: author-model (Codex), no
+  independent human review. Accept whatever the located code and tests show;
+  a missing guard is evidence of an unverified path, not a proven defect.
+
+## Outcome (complete after execution)
+
+Pending.
+
+Append an entry to research/evolution/ledger.json with `recording: prospective` and
+`protocol_ref` pointing here. Update profiles, run evolution checks, then snapshot at
+the agreed review boundary. No automatic maturity promotion.