Kernel panics, probably temperature related: Replace the heat sink?
TL;DR: Replace heatsink, or pay Apple for new logic board, or return to Amazon?
A couple weeks ago I bought an allegedly-refurbished 2013 15" Retina Macbook Pro (16GB/512GB, dual graphics) on Amazon.
Everything seemed hunky dory until Tuesday of this week, when it started giving kernel panics left and right. Once it starts panicing, it keeps panicing. I find that when it is cool, I'm able to get some use out of it. I put it in the cold garage in order to wipe my files off in preparation for a possible return or trip to the Apple Store, and was able to do that and also was able to reinstall Sierra. I took it back upstairs and a little while later it became unstable again.
Installing iStat Menus and cranking up the fans helps significantly, but only for an hour or so. The temps it shows aren't particularly high. If I start seeing temps around 55C or above, it starts crashing soon after.
I tried a RAM test tool but it doesn't find any RAM errors before the laptop crashes.
One possibly significant item: A couple days before this started I began using the laptop connected to my 27" cinema display, using both displays. I suspect that would automatically make it use the NVidia GPU, whereas it was probably using the integrated GPU before that. (I am seeing kernel panics regardless of which GPU is in use.) My thinking is that using the NVidia GPU may have caused heating which broke or weakened some shoddy soldering, resulting in the instability since.
Given the temperature sensitivity, and how keeping it cool prolongs normal operation, I'm pretty sure it's a thermal problem with either the CPU/GPU solder connections, or the heat sink and thermal compound.
So I see iFixit with the instructions on how to replace the heat sink. My question is, how would I know whether this would be worth trying?
If the problem is that the NVidia needs reflowing, then I don't think the heat sink would help. And I'm not sure if I want to try doing that myself.
I'm trying to decide should I return the laptop, or should I pay Apple $300+ for a depot repair of the motherboard, or should I pay $60+ to try replacing the heatsink and perhaps fixing any badly applied thermal paste.
If I do the latter, would it make it harder to get Apple to do a depot repair?
I've ordered a pentalobe driver from iFixit and some thermal compound & cleaner. When it gets here, I'm going to take a look inside. Assuming I don't see anything like Mountain Dew residue, I'm going to clean the fans if necessary, check that the SSD card is seated correctly (many of the panics I've gotten recently have been in code related to HFS, fsck, etc), and redo the thermal compound on the heat sink.
If that doesn't help, I'll likely return it.
Is this a good question?