docs(release): v0.21.0 notes should only cover its own delta

Duplicated the entire v0.20.1 fix bundle verbatim instead of
following house convention (each vX.Y.Z.md covers only what changed
since the immediately preceding release file — see v0.19.1 -> v0.19.2,
neither restates the other). v0.20.1 is a real, already-tagged
release with its own notes file; 0.21.0 only needed to add the
monitor-streaming feature, the KDE registry fix found while building
it, and the CI-only winget fix, with a pointer back to v0.20.1 for
the rest.

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

docs/releases/v0.21.0.md

@@ -1,5 +1,7 @@
 Update whenever it suits you — the app and the machine you stream from can be updated one at a time, and everything already paired keeps working.
 
+The headline: on Linux, Punktfunk's host can now stream one of your machine's own physical monitors instead of always creating a virtual one. This release also carries a KDE Plasma fix found while building it: on Plasma 6.7 and newer, the host had silently been falling back to a slower, less reliable way of arranging displays on every single session.
+
 ## New: stream one of your actual monitors, not just a virtual screen
 
 Until now, Punktfunk's Linux host always created its own virtual display to stream from — even on a machine you're sitting in front of, with real monitors already showing something. You can now point it at one of your machine's real monitors instead: pick it from a new **Streamed screen** card in the console, or pin it once in the host's configuration for an unattended box.
@@ -14,44 +16,6 @@ For headless and unattended setups, two new commands help you find and verify th
 
 On current KDE Plasma (6.7 and newer), Punktfunk's direct way of talking to the desktop about display layout stopped seeing any displays at all, with no visible error — every session quietly fell back to an external helper tool that's known to hang under load, exactly the thing the previous release's Plasma fix was meant to stop relying on. Fixed: Punktfunk now recognizes both the way older and newer Plasma releases announce their displays.
 
-## Fixed: installing or updating the Windows host
-
-Three separate reports came in within hours of the last release, all blocking install:
-
-- **`winget install` could fail outright if you'd ever had another streaming host on the machine** — even one you had since disabled or mostly uninstalled. A leftover, disabled service or an empty folder from an old install was enough to trip the "conflicting host" check and abort. Only a host actually set to start on its own counts now.
-- **Installing through a package-manager app (like UniGetUI) could fail with a file-path error** before the install wizard ever appeared, because of a broken setting the installer passed itself. Fixed.
-- **The package source instructions were incomplete.** `winget install unom.PunktfunkHost` says "no package found" until you've pointed winget at Punktfunk's own source, once per machine — the command for that is now spelled out.
-
-## Fixed: Moonlight compatibility no longer turns on without asking
-
-A fresh Windows install used to enable Moonlight-app compatibility silently, with no checkbox ever shown — even though that mode pairs over a less secure connection and the host warns about it on every start. It's opt-in now. Existing installs are untouched either way; Punktfunk's own clients were never affected.
-
-## Fixed: switching into Steam's Game Mode on Linux could black out the machine
-
-The previous release shipped the permission needed to properly take over the screen for Steam's Game Mode — and surfaced a few sharp edges that only show up once that takeover is actually live:
-
-- **The takeover could kill Punktfunk's own background service a few seconds later**, leaving the display dark with nothing left to bring it back except a physical console or SSH.
-- **It could also hang waiting on a permission prompt nobody could see**, delaying recovery until well after the stream had already dropped — or act on a leftover marker file from months ago and jump the machine to the desktop for no reason.
-- **Restarting the host mid-stream** — including a routine update — **could leave the display stuck off**, since nothing told the screen to come back.
-
-All three are fixed. Recognizing the machine's built-in Steam Game Mode may now need a one-time permission on some distributions; installing or updating normally takes care of it.
-
-## Fixed: a laptop with its built-in screen turned off stuttered constantly
-
-On a laptop with two graphics chips and the internal screen switched off — a common setup for streaming from a laptop docked to a TV or monitor — streaming stuttered roughly every two seconds, for the whole session. Punktfunk now recognizes a dark-but-connected laptop screen as the cause and points at the actual fix, instead of pointing at an external monitor that isn't the problem.
-
-## Fixed: latency creeping up over the course of a fast, high-bitrate stream
-
-On a fast local network, especially at higher bitrates, latency could climb steadily over a long session instead of staying flat. Several separate causes are fixed at once: the stream's internal clock sync could get starved by a busy connection and let the two ends' clocks drift; a stream mode could put more data on the wire than its bitrate setting actually allowed; a slow-to-decode moment could build up a backlog that took a while to work through instead of catching up immediately; and the one-time startup speed test could be mistaken for real network congestion and permanently cap the stream below what the connection could actually handle.
-
-## Fixed: missing cover art looked broken
-
-A library entry with no cover art showed as an empty grey box, which read as something that failed to load. It now shows a controller icon instead, so "no art yet" looks intentional rather than broken.
-
-## Improved: the Windows app now keeps its own log file
-
-The desktop app writes its own log to `%LOCALAPPDATA%\punktfunk\logs\client.log` now, instead of only printing to a console window that a normal launch never shows. If you're ever asked to send logs for a problem, this is where they are.
-
 ## Under the hood (for developers)
 
 - **Nothing changed on the wire.** Streaming protocol stays at version **2**, the embeddable C ABI at **13**, the Windows virtual-display driver protocol at **6** — 0.18–0.21 hosts and clients keep mixing freely. `GET /display/monitors` is a new endpoint (always 200, an explained empty list on a compositor-less host or a nested gamescope session); the display policy gains an optional `capture_monitor` field, resolved env-over-policy so a host.env pin outranks a console change. Both are additive.
@@ -61,14 +25,6 @@ The desktop app writes its own log to `%LOCALAPPDATA%\punktfunk\logs\client.log`
 - **libei absolute-coordinate resolution changed from matching by mode size to `mapping_id` → origin → size → first**, since two outputs can share a size but never a top-left, and a mirrored head's region is not the client's stream size at all. A named anchor that matches nothing falls back down the ladder rather than stranding input, and both a miss and a match now log once per distinct answer. `punktfunk-host anchor-test` exercises the ladder against a live compositor with two same-sized outputs — the one case a unit test can only simulate.
 - **`SWAYSOCK` is now derived rather than required to be inherited** — by the compositor's known PID, then the newest socket owned by the host's user — closing the last session variable a `systemd --user` host lacked for sway. A new drop-in (`PartOf`/`WantedBy` on `graphical-session.target`, opt-in, shipped under `/usr/share`) restarts a desktop-login host with its desktop, so a Wayland socket and portal connection that die with a Plasma/GNOME restart don't leave the daemon silently unable to recover.
 - **KWin's in-process output-management path now also binds `kde_output_device_registry_v2`**, not just the per-output `kde_output_device_v2` globals it originally shipped with — KWin 6.7 stopped advertising the latter, so the module written specifically to avoid shelling out to `kscreen-doctor` (0.19.x) saw zero devices and silently degraded to it on every session. Both models are supported now; registry-sourced devices arrive one Wayland round-trip later, so the handshake gains one conditional extra barrier.
-- **Windows conflict detection** now keys on the competing service's `Start` registry value (0–2 = boot/system/auto only), matching the tray's own dormant-is-harmless logic from last release; install directories no longer count.
-- **The winget `Log` switch used the literal string `|LOGPATH|`** instead of winget's actual `<LOGPATH>` substitution token, so Inno received an illegal filename on any install that requested a log (UniGetUI does by default). Guarded by a new manifest check.
-- **`gamestream` installer task lost its default-checked flag**; a silent/unattended install now lands `PUNKTFUNK_HOST_CMD=serve`. `/MERGETASKS=gamestream` opts in. Fresh installs only — `GamestreamParam` already omitted the flag on upgrades.
-- **The gamescope display-manager takeover** now calls privileged `systemctl` verbs non-interactively (`--no-ask-password`) so a headless polkit prompt can no longer block the stream's own recovery thread; `session_select_requested`'s sentinel baseline is now recorded at takeover time as well as at successful launch, closing a window where a months-old `~/.config/steamos-session-select` write could false-positive a session-select honor; and `punktfunk-host` now handles SIGTERM/SIGINT by restoring the display manager it stopped (20 s grace) before exiting, so a `systemctl --user restart` or package upgrade mid-takeover can't strand the box. The takeover also now ensures the invoking user has systemd lingering enabled before touching the display manager — without it, stopping the DM stops the user's own service manager ~10 s later and takes the host down with it.
-- **A new `internal_panel` target class** lets the exclusive-topology stall detector name a deactivated eDP/LVDS laptop panel as a disturbance suspect; previously the suspect list was filtered to external physicals only, so the one display responsible on a hybrid laptop reported as none. `EvtIddCxMonitorI2CTransmit/Receive` now answer every DDC/CI probe against the virtual monitor with an immediate `STATUS_NOT_SUPPORTED`, so third-party brightness/monitor tools stop occupying win32k's serialized physical-monitor path on it. The swap-chain drain thread now falls back to `TIME_CRITICAL` priority if MMCSS registration is refused, and raises `IddMinimumVersionRequired` 4 → 10 to use `IddCxSetRealtimeGPUPriority`.
-- **Clock re-sync** now spaces its probe rounds ~7 ms apart, tracks a running best-RTT guard band instead of a static connect-time floor, and applies the best batch of a rejection streak after three consecutive rejections rather than drifting unbounded.
-- **A new per-frame `WireBudget`** (shared between encoder backends) tracks the real bitstream-to-wire inflation ratio for datagram-aligned PyroWave sessions and deflates the rate-control target by it, so the configured bitrate pin holds on the wire rather than on the raw codec bitstream.
-- **The pre-decode frame channel now drains to the newest AU for all-intra (PyroWave) streams**, since those have no reference chain forbidding a mid-stream drop — capping any standing backlog at roughly one frame instead of ratcheting between the coarse jump-to-live thresholds.
-- **The ABR capacity probe's first post-probe report window is now discarded outright**, rather than measured, so the probe's own deliberate overload can no longer read as session congestion and permanently end slow-start early.
-- **New CI-only fix:** the winget release-verification step's `envs:` allow-list was a step-level sibling of `with:`/`env:` instead of nested inside `with:`, so `appleboy/ssh-action` never actually received it as an input and the step kept failing on every tag. Moved to match how `REGISTRY_TOKEN` is already forwarded elsewhere.
-- **New dev tooling:** `display-disturb` reproduces DDC/CI and modeset-class display disturbances on demand for stall-immunity testing; `punktfunk-host anchor-test`/`mirror-test` are described above. None of this ships to end users.
+- **CI-only fix:** the winget release-verification step's `envs:` allow-list was a step-level sibling of `with:`/`env:` instead of nested inside `with:`, so `appleboy/ssh-action` never actually received it as an input and the step kept failing on every tag. Moved to match how `REGISTRY_TOKEN` is already forwarded elsewhere.
+
+See [v0.20.1](v0.20.1.md) for the fixes carried by that release (Windows install/winget, GameStream opt-in default, the gamescope Game Mode takeover hardening, a laptop-panel stall fix, and the PyroWave latency-creep bundle) — all included here too, since this release supersedes it.