Skip to content

Physical display not restored after a Match Client session — Mac left with only a 1080p placeholder virtual display until reboot #8

Open

Description

@cnsren

Summary

After a remote Match Client session, the Mac's physical monitor didn't come back. The only display WindowServer knew about was a 1920×1080 placeholder virtual display. The real monitor showed only "Choose to mirror display from the menu". Reconnecting, replugging the monitor and sleeping/waking the display didn't help. A reboot fixed it.

Setup

  • Client: rootshell 1.0.16 (178) on iPhone (TestFlight), so rootshell-vnc 0.1.39. Match Client.
  • Host: Mac Studio (Mac17,14, M5 Max), macOS 27.2.1 (26B131). Screen Sharing enabled.
  • Monitor: Samsung Odyssey G95NC, 7680×2160, connected via a USB‑C to DisplayPort cable. It's the only physical display.
  • Path: from outside, through a NAT/VPN (the High Performance: no video behind NAT/VPN — client never sends UDP first, so no NAT mapping is created #7 path). High Performance failed to get video there (UDP [UNREPLIED] at the gateway), the session fell back to Standard, and Match Client still created a rootshell Virtual Display (2048×3328) over Standard. That seems to be the new 0.1.39 behaviour where usesAppleVirtualDisplays no longer requires High Performance.

What happened

  1. The remote session with the virtual display worked over Standard.
  2. The session ended. I don't know exactly how: the app may have been backgrounded or killed rather than disconnected cleanly. No debug log was enabled, sorry.
  3. Back at the Mac, the G95NC showed only "Choose to mirror display from the menu".

State of the Mac afterwards

  • No Screen Sharing connections on 5900, TCP or UDP. rootshell Virtual Display was gone.
  • CGGetOnlineDisplayList and SLSGetDisplayList returned one display: vendor 0x756e6b6e ("unkn"), model 0x76697274 ("virt"), 1920×1080, main. Its ID changed from 44 to 46 after a display sleep/wake. The G95NC wasn't in either list.
  • The hardware side looked fine. In IORegistry the DCP connection mapping still had "ProductName"="Odyssey G95NC", MaxW 7680 / MaxH 2160, and the USB‑C port showed the DisplayPort transport active with the G95NC attached.
  • SSMenuAgent relaunched twice in this period (new PIDs at 04:57 and 05:13 UTC on 2026-11-15).

What didn't help

  • Power-cycling the monitor and replugging the cable.
  • System Settings → Displays: the G95NC wasn't listed.
  • pmset displaysleepnow and then waking.
  • Connecting again from the LAN with Remote Display. That showed the 1080p placeholder desktop, and the G95NC still showed the mirror message.

What fixed it

A reboot. Afterwards the G95NC was back at 7680×2160 as the main display. Logging out would probably have worked too, since it restarts WindowServer, but I didn't try it.

Thoughts

It looks like the physical display topology wasn't restored after the virtual display session. Part of this may be macOS, since the host should recover by itself once the session is gone. But the trigger was a Match Client session that fell back from High Performance to Standard and still used the Apple virtual display command, which is new in 0.1.39, so I wanted to flag it. A few things that might help on the client side:

  • Explicitly restore the physical topology before disconnecting, and when going to the background.
  • Make sure a virtual display created during a session that's falling back is torn down before the Standard session takes over.

If it happens again I'll enable VNC Debug Logging and attach the log. Happy to test anything.

Activity

  1. kitknox commented on Nov 15, 2026

    @kitknox
    Owner

    There is definitely weirdness when using match client combined with standard. The official macOS client doesn't offer this at all. In fact even the high performance UDP mode in match client mode is buggy even with the macOS official client as soon as you have more than one display. I run by default my MacBook Pro in clamshell connected to Studio Display. If I connect with the official client in match mode and then disconnect, the display flips to extend mode thinking the internal display is active. The only way to recover is to open the lid and close it.

    For standard mode in my testing the virtual display doesn't close right away and takes ~15 seconds to close. I'll see if I can synthesize something to recover. I wonder if flipping back to a high performance mode connection and then disconnecting that recovers?

  2. cnsren commented on Nov 15, 2026

    @cnsren
    Author

    Thanks! One difference that may matter: my host is a Mac Studio (no internal display, no lid). The only physical display is a Samsung G95NC on USB‑C→DP. So instead of flipping to extend mode, WindowServer ended up with no physical display at all, just the 1080p placeholder (vendor "unkn" / model "virt"), and there's no lid trick to force re-detection. Only logout/reboot helped.

    On your idea: while it was stuck I did connect again from the LAN with Remote Display and then disconnected, and it did not recover. I didn't confirm whether that session ran in High Performance or Standard, though (most likely HP, since it was on the same subnet). If it happens again I'll test an explicit High Performance connect/disconnect with VNC debug logging on and report back.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions