Some clarification and reinstating of facts:
- SB default power settings: While on Battery, sleep at 10 minute mark and hibernate at 2 hour mark. While Plugged, sleep at 10 minute mark, and no hibernate at all.
- SB does not crash while in hibernate.
Almost none of the Win10 background process run while in hibernate, therefore no crashes take place.
If you leave SB on battery, the chances of getting a crash is minimum or nonexistent, because after 2 hours it will go into hibernate.
- SB does crash while in sleep (long period).
Almost all of the Win10 background process run while in sleep, therefore crashes take place.
If you leave SB plugged, the chances of getting a crash is very very high if SB stays in sleep for long period of time (several hours) or overnight.
- SB firmware got updated first time on 2/17/2016 which modified the sleep state in general (the default power settings did not change).
My SB sleep-crashes started after the 2/17/2016 firmware update. Before the 2/17/2016 firmware update, I never experienced sleep crashes (plugged in) since I got my SB on November 28.
I have concluded that Win10 background processes are the culprit. They do not work properly with the modified sleep state (rolled out on 2/17). The EventViewer (under Windows Logs, System) shows many critical errors while SB is in sleep, most are DCOM errors.
Eventually, while sleep, Win10 crashes.
A very common DCOM error (many many of them) while in sleep, is related to InsiderHub and FeedbackApp.
My experience is the same.
I have not changed the default power settings. If I plug in my SB overnight to recharge then it almost always crashes, whereas if I leave it unplugged -- as I frequently do now, for this very reason -- then it almost never crashes. I can't say 100% that
it's always in the one case and never in the other but the difference is clear.
I didn't get my SB in the UK until Feb 17 but it didn't start crashing immediately: the first occurrence in event log was on the 23rd, which was a couple of days after the 17/02 firmware update was applied (it didn't come with that, obviously) so that could
be the culprit.