diff --git a/work/projects/cpm-status-2026-10-15.md b/work/projects/cpm-status-2026-10-15.md
index 7916a814..fb71a59a 100755
--- a/work/projects/cpm-status-2026-10-15.md
+++ b/work/projects/cpm-status-2026-10-15.md
@@ -583,3 +583,10 @@ Known-failures suite начат с включённым expensive flag. Перв
 Первая попытка attach до idle изменила условия: idle-helper не дождался выгрузки за 45 с; этот debugger run — отдельный infrastructure failure, не результат P5. Attach перенесён только после двух idle cycles и window close.
 
 `pinned-lldb-after-idle.log`: breakpoint в `Common.breakByRaisingSigInt`, stack `NSObject.DeallocationCheckAction.interrupt → ensureObjectDeallocated`, parameter description = **MainWindow has not been deallocated**, line NSObjectExtension.swift:80. Это уточняет прежнюю гипотезу: controller уже прошёл deinit, сработала следующая 8-секундная проверка его окна. Исправления только harness: создание NSWindow/MainWindowController целиком в отдельном autoreleasepool (как в существующих DeallocationTests), lookup/selection не держит controllers через await, closing window немедленно удаляется из owned list. Guards приложения и production window-close path не отключаются. Нужен повтор без debugger; прежние app-level P5 результаты пока red/incomplete, хотя actual first failure/tab-not-found уже получены.
+
+
+### 16 Oct: signed cache/P5 follow-up (intermediate, not a green suite)
+
+- `signed-pinned-autorelease.xcresult`: real shared pin + two normal WebKit idle unloads + real window closure again produced `runtime.sendMessage(): Tab not found`, missing page initResp, production timeout `init_stage=none`, `tab_in_context=false`, matching controller and `bg_view_creations=2`. The host still interrupted before the second probe/attachment. Prior LLDB identifies the DEBUG deallocation guard for **MainWindow**, not an assertion from the CPM detector. Wrapping creation/selection in autorelease pools alone did not resolve that retention; the test is still red.
+- `signed-large-cache.xcresult`: fresh persistent UUID store, **20,000** cacheable loopback resources, actual disk-cache records, fill **42.705 s**. Real extension reload changed baseURL; immediate next navigation succeeded in **1.058 s**, page initResp present, background created once and ready under 1 s. `p3NoBackgroundFailure=false`: **P3 not reproduced in this run**. The readiness control also succeeded. The test failed only in teardown (`WKWebSiteDataStore: Data store is in use (by network process)`), while its own controller/store references were still retained. Cleanup is being corrected before a new run.
+- These runs use the project's normal **Apple Development** signing. The earlier unsigned CLI workaround was incorrect and removed; no SecureVault bypass remains. Current verified distinct passing methods remain **18/20**, with the two red methods above; there is no combined all-green result.