Why UEFI works and Windows doesn't:
UEFI only needs a simple, fixed-mode framebuffer. When Windows takes over, the Intel driver has to reprogram the display engine, retrain the eDP link to the panel, and re-enable power features such as PSR, backlight PWM and the display low-power firmware. Any step that fails leaves you with a black screen, corruption, or a fallback to 1024×768 on the Basic Display Adapter, while the rest of the system boots fine. That is why you hear the startup sound.
Why a cold boot can fix it:
A normal Restart or Shutdown → Power On does not necessarily produce exactly the same hardware state as removing power completely.
A forced shutdown followed by a cold start can reset:
Intel GPU/display-engine state
eDP link state
panel power state
graphics-driver state
ACPI display power state
Modern Standby-related state
The next boot therefore gets a cleaner graphics initialization, allowing the internal display to come back.
What the underlying fault probably is. There are two candidates:
A software/firmware bug in panel power-state handling, for example a PSR or low-power state that the driver and Surface firmware disagree about. This is plausible, but it rarely causes visible glitches or distortion.
Marginal hardware in the display flex cable or connector, the panel's timing controller, or the GPU/board. Flicker and distortion that depend on thermal and power state fit this better, and a link that trains badly on a warm restart but fine when cold is the classic signature.
The odd 1736×1824 active signal, if it was read exactly (the native panel is 2736×1824), supports a link or timing negotiation problem rather than a simple driver crash.
I can't prove which one it is from here. The quickest way to tell is to wait for a black screen and then try three things:
#Win+Ctrl+Shift+B. If the screen returns, it's a driver or WDDM issue.
#An external monitor over USB-C. If it works while the internal panel stays dead, the fault is in the panel path. If it glitches too, suspect the GPU or driver.
#Disabling the Intel adapter so Windows uses the Basic Display Adapter. If the failures continue, the Intel driver isn't the cause.
If the failures survive a clean install of Microsoft's Surface driver package and the Basic Display Adapter test, I would treat it as hardware.