Bug 2389415

When using QXL + SPICE on Windows guests with display scaling set to 125% or 150%, the UI freezes after a few minutes of normal use. The VM remains alive on the network (ARP/RDP responds), so it looks like a display path freeze, not a full guest hang. Switching the display adapter to virtio(-vga)/virtio-gpu avoids the issue but loses QXL auto-resize behavior. This appears related to the QXL/cursor path regression discussed upstream for 9.1.x and fixed in the 10.0.x series. Environment: Fedora 42, qemu-9.2.4-1.fc42, libvirt/virt-manager, Windows 10 22H2 and Windows 11 24H2 guests, spice-vdagent installed in the guest.

Debian bug: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1084199
QEMU issue: https://gitlab.com/qemu-project/qemu/-/issues/1628

Reproducible: Always

Steps to Reproduce:
1 - Create or run a VM with Video set to QXL and SPICE display.
2 - In the Windows guest, set display scaling to 125% or 150%.
3 - Use the system (Explorer, browser). After some minutes, the guest UI stops updating and input no longer affects the on-screen UI, while the VM continues to respond on the network.
Actual Results:
The guest UI freezes after a short period; CPU usage stays low; libvirt reports the domain as running; the VM is reachable over the network (e.g., via RDP/ARP).

Expected Results:
No freezes when using QXL with 125–150% scaling; display should remain responsive.

Additional Information:
Package/version: qemu-9.2.4-1.fc42 (Fedora 42).
Workaround: switch the display to virtio(-vga)/virtio-gpu or access the guest via RDP.
Upstream references for the regression/fix:

Debian bug 1084199 (marked fixed in qemu 1:10.0.0~rc2+ds-1): https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1084199

QEMU upstream issue 1628 (cursor/graphics stream desync discussion): https://gitlab.com/qemu-project/qemu/-/issues/1628

Request: please backport the relevant QXL fix from 10.0.x to Fedora 42 (or temporarily revert the problematic change). I can test updates-testing builds.