LrC 15.4.1 / 15.1 / 14.5.2 - 100% reproducible crash (ntdll heap corruption) importing >100MP images; disabling GPU has no effect

SUMMARY
=======

Importing a large image (approx. 143 MP) into Lightroom Classic crashes the
application 100% of the time on my machine. The same crash occurs on three
different LrC versions, including after a full downgrade, so it does not appear
to be specific to one build. Smaller images import fine, so this is a
size-dependent failure in the import path.

CRASH SIGNATURE (identical across 10+ occurrences)
==================================================

Faulting application : Lightroom.exe
Faulting module      : ntdll.dll (10.0.19041.7548)
Exception code       : 0xc0000374 (STATUS_HEAP_CORRUPTION)
                       0xc0000005 (ACCESS_VIOLATION) on some occurrences
Exception offset     : 0x00000000000ff509  <-- identical every single time
WER Sig[7]           : PCH_08_FROM_ntdll+0x000000000009E164

Adobe crash reporter stack, top frames:

  [0] mems...  (memset / memcpy family)
  [1] RtlAllocateHeap
  [2] RtlAllocateHeap
  exception = EXCEPTION_ACCESS_VIOLATION

Windows subsequently applied FaultTolerantHeap to Lightroom.exe (marker ffffbaad).

STEPS TO REPRODUCE
==================

1. Import an image of approximately 10192 x 14016 px (about 143 MP) into a catalog.
2. Lightroom hangs (hourglass) and then crashes, or crashes immediately.

Occurs both from the Import dialog and via drag-and-drop.

TEST MATRIX (same machine, same session conditions)
===================================================

OK     3.0 MP    2000 x 1500   JPEG, sRGB ICC          re-encoded via Python/PIL
OK    35.7 MP    5096 x 7008   JPEG q97, sRGB ICC      re-encoded via Python/PIL
CRASH 35.8 MP    5100 x 7014   TIFF, uncompressed      scanner output (WIA), no ICC
CRASH 35.8 MP    5100 x 7014   TIFF, LZW, sRGB ICC     re-encoded via Python/PIL
CRASH 142.9 MP  10192 x 14016  JPEG, no ICC            scanner output
CRASH 142.9 MP  10192 x 14016  JPEG q97, sRGB ICC      re-encoded via Python/PIL

Notes:
- Re-encoding through a clean, standards-compliant writer does NOT avoid the
  crash at large sizes, so malformed scanner output is not the cause.
- TIFF appears to fail at a lower pixel threshold than JPEG.

THINGS RULED OUT (each verified, not assumed)
=============================================

System RAM exhaustion
  17-18 GB free at the time of every crash (31.4 GB total). Pagefile peak usage
  only 359 MB of 8.7 GB allocated. Commit limit 39.9 GB with 21.9 GB free.

Faulty RAM
  Windows Memory Diagnostic, full standard pass over all 32,160 MB: Pass, 0 errors.

GPU acceleration
  Disabled in Preferences and verified on disk:
    GPUManagerPref = "off"
    useGPUForPreviews = "off"
    useGPUCompute = false
    useGPUForExport = false
    useGPUInLibrary = false
  Crash persists.

  IMPORTANT: with all of the above set to off, the Camera Raw log still shows
  262 lines of "Building compute pipeline state" and still reports
  "GPU4 Soft Status Result: fail_insufficient_memory".
  GPU initialisation does not appear to be fully skipped when GPU is disabled.
  This may be relevant to the root cause.

AI / Assisted Culling
  AICulling_enableAutoCullingAnalysis = false, set directly in the preferences
  file. Crash persists.

Camera Raw cache
  Purged via Preferences > Performance. Crash persists.

Catalog corruption
  PRAGMA integrity_check on the .lrcat returns ok.
  A brand-new empty catalog in a different location also crashes.

Preview cache corruption
  previews.db and root-pixels.db integrity checks return ok.

Application version
  Reproduced on 15.4.1.1, 15.1.0.5 and 14.5.2.1.

Reinstall
  Clean reinstall performed, no effect.

Third-party DLL injection
  All 249 modules loaded into Lightroom.exe are Adobe or Microsoft.
  AppInit_DLLs is empty.

Plug-ins
  No third-party plug-ins installed (Modules folder does not exist).

File integrity
  JPEG SOI/EOI markers and TIFF headers verified intact.

Antivirus
  Only Windows Defender is active (McAfee is installed but its service is stopped).

GPU driver timeout
  No TDR or display reset events in the System log.

ADDITIONAL CONTEXT
==================

- No Windows updates, driver updates or software installs occurred on the day
  the crashes began. The System event log contains no errors or warnings prior
  to the first crash.
- Editing already-imported photos worked normally for hours immediately before
  the first crash. The crashes began only when large scans were first imported.
- explorer.exe also crashed twice on the same day, in different modules
  (DUI70.dll and combase.dll).

ENVIRONMENT
===========

Lightroom Classic : 15.4.1.1 (also reproduced on 15.1.0.5 and 14.5.2.1)
OS                : Windows 10 Home 22H2, build 19045.7548
CPU               : AMD Ryzen 9 5900HS
RAM               : 31.4 GB
Machine           : ASUS ROG Flow X13 GV301QE (hybrid graphics, no MUX switch)
iGPU              : AMD Radeon Graphics, driver 31.0.21912.14 (2024-02-20), 496 MB dedicated VRAM
dGPU              : NVIDIA GeForce RTX 3050 Ti Laptop, driver 32.0.16.1088 (2026-07-22), 4 GB VRAM
Image source      : HP DeskJet 4720 flatbed via Windows Fax and Scan (WIA), 1200 dpi A4

ATTACHMENTS AVAILABLE ON REQUEST
================================

- 10 crash dumps (Lightroom.exe.*.dmp, approx. 60 MB each)
- Camera Raw log showing GPU init and fail_insufficient_memory
- Windows Application event log entries for all crashes

CURRENT WORKAROUND
==================

Downscaling the scans to approximately 35 MP before import avoids the crash entirely.