docs(windows-host-goal1): record §2.5 done + on-glass results + Remaining list

The plan tracker referenced "§2.5 — see below" but had no §2.5 section and no "what's left". Add:
  * a Status banner (all 6 stages + §2.5 done; branch not merged),
  * the §2.5 section — the 3-step ownership-model rewrite (VirtualDisplayManager/MonitorLease,
    the deleted globals), the CURRENT_MON_GEN-write-only finding, and the on-glass reconnect-leak
    result (the vdm-init-order panic found+fixed, 0 leaks, IDD-push zero-copy verified),
  * a "Remaining (next session)" list: EncoderCaps, optional namespace collapse, merge to main, and
    the pf-vdisplay driver slot-reclaim fix (driver WIP, not the host refactor) with the dev scripts.
Mark §2.5 IMPLEMENTED in the design doc (windows-host-rewrite.md) with the write-only-gen deviation.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

docs/windows-host-rewrite.md

@@ -229,6 +229,12 @@ path readable, and removes the capture/encode backend-disagreement bug class. It
 
 ### 2.5 Ownership model — delete the global statics
 
+> **✅ IMPLEMENTED (2026-06-25, branch `windows-host-goal1`).** Landed as 3 steps + an on-glass
+> reconnect-leak test — see [`windows-host-goal1-plan.md`](windows-host-goal1-plan.md) §2.5 for the
+> commits + results. One deviation from the sketch below: the 5-agent map found **`CURRENT_MON_GEN` was
+> write-only** (the per-frame monitor-gen bail was never wired), so the "generation carried through
+> `WinCaptureTarget`" item was unnecessary and dropped; the gen lives on the manager + lease only.
+
 Today the lifecycle is smeared across `IDD_PERSIST` + `open_or_reuse` (dead code), `CURRENT_MON_GEN`
 (read per-frame), `IDD_SETUP_LOCK`/`IDD_SESSION_STOP` (the preempt dance), `MGR: Mutex<Mgr>`, and on
 the driver side `ADAPTER`/`MONITOR_MODES`/`NEXT_ID`/`WATCHDOG_*`/`DEVICE_POOL`. Replace with: