Record original room requirement and cancelled reservation scope gap

.project-docs/30-worklog/tasks/20261008-production-review-9e7b.md

@@ -85,3 +85,12 @@ Read: memory-index, project-positioning, current-state latest September sections
 - Verification:74 Python tests passed without skips (40 backend/config/local-wrapper tests and34 existing source-review/runtime tests, disposable local PostgreSQL enabled),31 JavaScript tests passed, node syntax and Git whitespace checks passed. Tests cover two pending days, same-day deduplication, queue restart, server list restoration, per-request drafts, lost new submission and reopening a completed review.
 - Local activation: gracefully restarted the existing private local service only after verifying zero active queued/downloading/processing tasks. No actual business value was changed and no new Oracle request was made. CUA verified choosing9/17 enables “下载并处理” while10/7 still shows0/26; the pending-date entry returns to10/7 unchanged. The page is left with9/17 selected for the user's explicit submission. Screenshot: private `production-validation-20261007/independent-report-dates-20261008.png`.
 - Promotion candidate: document independent report-date progress and the persistent pending-field-review entry in canonical product flow at integration. This preserves existing business rules and user-approved date interaction; no policy reversal is involved.
+
+## Same-task Follow-up: Existing Room Requirement
+
+- User asks whether room validation existed before the new field-review page. Same-task ownership was resumed and verified; this is a read-only rule/history review, with no change to product behavior.
+- Evidence: the earliest repository commit `a701de9` (2026-07-29) already required nonempty `DISP_ROOM_NO` for whitelist candidates (`XML_ROOM_MISSING`) and deduplicated by room plus arrival date. Current locations are `process_daily.py:995` and `:1131`; Git blame traces both to that initial commit. This is a nonempty-field check, not a lookup verifying that the hotel has that room. Original XML missing-room errors failed the batch; the new upstream field-review page makes that existing problem editable before processing.
+- The original XML processor does not read reservation status and does not exclude Cancelled/NoShow; it filters by rate whitelist, validates fields and deduplicates. This does not establish Oracle's manual ARR report selection defaults or prove that cancelled reservations belong in ARR.
+- Saved10/7 evidence for source row1 / confirmation300007418 explicitly has `reservationStatus=Cancelled` and no identified room (`calendar_missing_rooms`). The previous user-facing instruction to fill this room was incomplete: first determine whether such a reservation should enter the report. Do not fabricate a room or silently introduce status filtering. Raw source and review decisions remain unchanged.
+- Verification: read existing processor/independent-validator/contracts, inspect immutable saved reservation response, and independently review baseline Git history; no network reads, test runs, service changes or real manual decisions were needed.
+- Follow-up/promotion candidate: settle the upstream ARR selection policy for cancelled/unassigned reservations against original report settings or an explicit business decision before changing inclusion, allowing a blank required room, or changing deduplication. Current source evidence alone cannot resolve that policy. Explain the existing requirement separately from the open report-scope question.