Lenovo Legion Y540 shuts down on some games

My laptop has crashed several times since October 31, 2022. I read a lot of threads on various websites (including this one, Reddit, Quora, Microsoft forums, and Tom's Hardware), but I couldn't find an answer that would help in my situation.

It's possible that these problems began shortly after I replaced the thermal paste, but I'm not certain. I think I replaced it because these problems started, but I'm not sure what was the actual order. Still, it's something to think about. I replaced it again today, but this time using a smaller amount of paste. It didn't help.

Here's a link to my Speccy snapshot: http://speccy.piriform.com/results/qFrmWhTOVIT0A0o2hPUSKNx

The crash happens like this:

1. Start a game (it doesn't appear to happen in all games).
2. Play for an unspecified amount of time (sometimes it happens right after launch, sometimes I'm able to play for a couple of hours).
3. The screen goes black without warning, the fans go mental, and the laptop shuts down after a few seconds. I might as well have the game paused in the background and read something in the web browser for the crash to occur. However, the game must be running.

I've already tried some of the more common suggestions from other threads:

1. Testing different NVIDIA drivers. Installing different drivers has no effect on whether or not these crashes occur, but it does alter which errors are logged in the Event Viewer.
   Every test was performed after cleaning up the previous driver with DDU in the Safe Mode. I tested the most recent driver, the one provided by Lenovo for this specific laptop, and at least two other versions.
   • For newer drivers, the source of the crash is "nvlddmkm." There are multiple logs of Event ID 14 when the screen goes black. No helpful data is available for these events, except for their ID.
   • For older drivers, the source of the crash is "dwm.exe" with Event ID 1000 (see: Report-DWM.wer.txt for an error report extracted from C:\ProgramData\Microsoft\Windows\WER\ReportArchive). Other exes appear to be logged as sources at times, so I don't believe dwm.exe is the root of the problem.
   • Despite having multiple crashes, only one produced a Minidump (see: 112722-10812-01.dmp).
2. Hardware stress-testing. This part is interesting, because stress-testing does not result in a black screen. I ran all options available in OCCT. I ran the "Extra" and "Ultra" stress-tests in 3DMark. I ran prime95 tests to check the RAM. Everything seems to be working just fine.
3. Replacing the charger. In fact, I was hoping it was the culprit. I had been using a "modified" charger for over a year. I broke the plug on the original one, so I soldered a cable from another charger that I had on hand. Every time I turned on the laptop, I received a warning that I was using a charger with insufficient power for my notebook. Lenovo plugs include a resistor that indicates the power of a given charger, so in my case it was a false warning. Because of this minor detail, I reasoned that my laptop might be more careful with its power consumption. The problem persisted even after I replaced it with a new 230W one.
4. Connecting the charger directly to a power outlet, instead of using an extension cord. No luck.
5. Slightly underclocking the GPU. I didn't overclock it before, so by underclocking I was changing the factory settings. It didn't help.

I'm also attaching a log file from the GPU-Z with sensor readings just before the crash (see: GPU-Z Sensor Log.txt). The crash in the log happened at 23:24:38, and produced a clear cut-off point in the log.

If you require any additional information, please let me know.
If something else comes to my mind, or I have new test results, I'll make sure to post them

I ran some more tests yesterday.
One of the games that crashed used a custom DirectX 11 renderer (https://github.com/Kirides/GD3D11). There were no more black screens after I removed the DX11 mod from the game files. I'm not sure if it was due to DX11 compatibility, or if removing it simply reduced the load on my laptop, but I don't think the mod itself was the cause. I say that because at least one other game crashes, and it is not modded.
Logically, it's either a software or a hardware issue. Please correct me if I'm wrong, but if it was a software problem (like the GPU driver), wouldn't it happen during the 3DMark stress tests, which used DX11 and DX12? Similarly, if it was a hardware issue, wouldn't it happen during stress testing?