Thank you for your reply; I've just tried that: clicking OK to load factory presets crashes blender.

Logs attached.

(I've updated to 3.4.1 before reading your reply, but no difference re: the errors and crash)

Starting Blender from blender_factory_startup.cmd doesn't produce any crash.
So, the problem is somewhere in the handling of one or more add-ons, but so far I couldn't track it down to one. If given the logs above there's something I could try, please advise, thanks.

[UPDATE]: basically at each addon I find referenced into the crash logs that I remove, Blender finds exception with another.

For instance:

Traceback (most recent call last):
  File "D:\+++ PROGRAMS +++\Blender Foundation\Blender 3.4\3.4\scripts\modules\addon_utils.py", line 421, in disable
    mod.unregister()
  File "C:\Users\marco\AppData\Roaming\Blender Foundation\Blender\3.4\scripts\addons\random_flow\__init__.py", line 8743, in unregister
    bpy.utils.unregister_class(cls)
RuntimeError: unregister_class(...):, missing bl_rna attribute from 'RNAMeta' instance (may not be registered)
Error   : EXCEPTION_ACCESS_VIOLATION
Address : 0x00007FF7D70195C5
Module  : blender.exe
Thread  : 0000847c
Switching to fully guarded memory allocator.

Hi, thanks for the update. I don't think there is any shortcut to find the culprit add-on
You'd need to disable them one by one to find the add-on that leads to crash.

RuntimeError: unregister_class(...):, missing bl_rna attribute from 'RNAMeta' instance (may not be registered)

Maybe unregister is called without registering a class.

Thank you, I will remove them one by one, and report to their devs, if I find anything.

Just to help me go about that, though, it's likely that the culprit is an actual addon, and not some kind of other degradation on the userpref file, (for instance, some path manually defined by me inside an addon's preferences, or something like that), correct?

Just as an update, I've removed more addons and the unregistering error has gone away, but Blender still crashes; in the debug output this error :

Exception ignored in: <function ImagePreviewCollection.__del__ at 0x0000021CDE9CF880>
Traceback (most recent call last):
  File "D:\+++ PROGRAMS +++\Blender Foundation\Blender 3.4\3.4\scripts\modules\bpy\utils\previews.py", line 64, in __del__
    raise ResourceWarning(
ResourceWarning: <ImagePreviewCollection id=0x21ce8032d90[0], <super: <class 'ImagePreviewCollection'>, <ImagePreviewCollection object>>>: left open, remove with 'bpy.utils.previews.remove()'
I1229 13:51:07.152995 36736 device.cpp:32] CUEW initialization succeeded
I1229 13:51:07.152995 36736 device.cpp:34] Found precompiled kernels
I1229 13:51:07.177768 36736 device.cpp:56] HIPEW initialization failed: Error opening HIP dynamic library
Error   : EXCEPTION_ACCESS_VIOLATION
Address : 0x00007FF7D70195C5
Module  : blender.exe
Thread  : 00008f80
Switching to fully guarded memory allocator.
Blender 3.4.1

Which seems to be related to cache size or limit: https://developer.blender.org/T93888

Since it does not happen when Blender is started without reading the prefs, I have to think that either a setting there, or a cache saved there (if any?) is called for?

What's odd is the HIP error with a RTX device, I think.

To test it, I've downloaded Blender 3.3, performed a clean install, (so, new prefs, startup, everything), and started it with debug.cmd, and I got the same HIP error, but no crash.

3.4.1 logs:

3.3.2 logs:

To be sure, here's the device list as outputted by Powershell "Get-PnpDevice"

I've installed 3.5 version too; vanilla install, first run and no addon:

• start;
• don't touch anything;
• edit prefs just to define logs location, nothing else;
• quit.

no crash, same HIP error as before.

• crash is therefore related likely to some addon combination or degraded prefs, class unregistering etc, or if not, it's not trackable.
• HIP error is worrying: why would a system with nvidia GPU use HIP library?

I've also installed version 2.93.13 to do teh same test as above; the result is different, with no HIP error.

Blender doesn't know you have an nvidia or amd gpu so it tries to load the cuda and hip libraries, if hip succeeds it knows there's an AMD gpu present, if cuda (listed as cuew in your log) succeeds i knows you have an nvidia gpu. judging by your crashlog the crash you are having is not related to hip rather by some python code or addon that has a bug.

2.93 doesn't show the hip error since 2.93 did not support hip and used opencl to support amd gpu's

Thank you very much for your explanation; if I can dismiss the HIP error in the log as pure info, and nothing else, then I will proceed to simply reinstall Blender 3.4.1, and then start to install the addons one by one, checking the log each time.