Rigid Body Crash when switching between MacOS and Windows

I’m a relative newbie on Blender 2.83 and have an infuriating problem when attempting to edit – on a Windows 10 laptop – an animation containing a rigid body simulation originally created on a Macbook Pro. The Blender file works fine under MacOS (Catalina) but when I try to make any changes involving the rigid body world under Windows 10, Blender crashes instantly with no error message or (as far as I can see) error log.

The reason I’d like to be able to switch between machines is that I prefer working on the Macbook but my Windows laptop has a more powerful GPU for rendering. Ideally I’d like to be able to switch between the two machines seamlessly.

I can happily edit all other aspects of the file and render individual images OK under Windows, but as soon as I run the rigid body simulation from the first frame it crashes instantly.

I’ve tried deleting the baked cache, but I note the rigid world panel is still reporting "240 frames in memory". I can find no way to remove these reported ‘in memory’ frames. I’ve also tried manually deleting the files in the temp directory, and assigning a disk cache location, but all to no avail.

It feels to me as though Blender is ‘remembering’ something about the rigid body world cache created under MacOS which I’m not able to ‘unremember’ or ‘reset’ under Windows.

Any tips as to how to resolve this? I can happily post the Blender file it that helps, but I suspect this is a generic problem.

EDIT Since posting the above, I’ve upgraded to Blender 2.83.4 but the problem still occurs. I’ve run in debug log mode and obtained the following output, for what it’s worth. This seems to indicate an OpenCL problem but I’m not clear how that’s relevant – my Windows laptop has an NVidia GTX1050Ti which does appear to be running in CUDA mode, and I can render individual images quite happily:

I0813 13:02:00.950862 6564 blender_python.cpp:191] Debug flags initialized to:
CPU flags:
AVX2 : True
AVX : True
SSE4.1 : True
SSE3 : True
SSE2 : True
BVH layout : BVH8
Split : False
CUDA flags:
Adaptive Compile : False
OptiX flags:
CUDA streams : 1
OpenCL flags:
Device type : ALL
Debug : False
Memory limit : 0
I0813 13:02:01.117919 6564 device_opencl.cpp:48] CLEW initialization succeeded.
OpenCL error: CL_INVALID_VALUE in __clewGetPlatformInfo(platform_id, 0x0904, sizeof(data), &data, NULL) (C:bbuildbot-worker-windowswindows_lts_283blender.gitinterncyclesdevicedevice_opencl.cpp:189)
OpenCL error: CL_INVALID_VALUE in __clewGetDeviceInfo(device_id, 0x1030, sizeof(data), &data, NULL) (C:bbuildbot-worker-windowswindows_lts_283blender.gitinterncyclesdevicedevice_opencl.cpp:210)
I0813 13:02:01.217612 6564 device_cuda.cpp:41] CUEW initialization succeeded
I0813 13:02:01.217612 6564 device_cuda.cpp:43] Found precompiled kernels