I'm curious what you'll find. The sane engineer side of me wants to think they have a faux crash somewhere to gather information on an interesting event. They pushed this update out to 15% of the fleet, for like a week before it rolled out to everyone, they had to have known the vision task crashes a lot.
so I just did some errands and that resulted in just one captured core, but it's truncate, the compressed size is ~3.5M, uncompressing it results in ~17M file, but when you load it in GDB, gdb says that the size is expected to be upwards of 294510592 bytes.
As such very little could be gleaned from this file other than it was caused by SIGABRT, so likely assertion of some sort.
I looked into how are those cores generated and essentially they have a core collector registered that takes the core, passes it through gzip -1 and writes output to /var.
now should the system power disappear of course the write would be incomplete. Interestingly I also have a lot of cores of zero size too that falls well into this picture if we assume that majority of the crashes are happening on or shortly before ape shutdown.
Now obviously I can only speak about my car, but given how external snapshot requests are firmware-independent and people associate this release with increased uploads only, the reason seems to be internal and so likely the core dumps. If a crash happens not before shutdown (or well enough before shutdown so all of it is written), the core seems to be quite big and as such would show up as a bigger uptick on the uploads graph. Or if you do a bunch of stops (doctor mode?) then I imagine even smaller partial dumps would accumulate to a significant amount.
I'll keep an eye on it in the coming days to see if anything else interesting arises.