A practical question to me is, is a USB stick installed Linux-BOINC client more or less resilient than one running off an SSD? Myself been running BOINC on a USB 3.0 64GB stick for the best part of a year and yet to encounter probs, bad sectors etc. That said, certainly have set BOINC to write little, longer intervals between checkpoints controlled via the Write To Disk setting [WTD]. That can be set, just confirmed again, to at least 999,999 seconds, i.e. just start job and have the final result output written to storage. Of course, that's a setting for hosts that have extreme reliable power-supply. Given that tasks generally do not take longer than 6-7 hours on modern devices with I5/I7, that's what you would loose at most, per task that's running. The WTD setting cannot control CEP2. That sciences writes up to 16 checkpoint per job, practically we've had a long series now doing 12 at most. What's wisdom, past tests showed there was no performance improvement setting WTD over 10-15 minutes, beyond which only wear is the criterion, if it is one.

Reminded me of another post recently where there was reports of enormous loads of write by BOINC to an SSD, and then it came out that utility was also logging the CPU L2 Cache transfer ;D

I like the write to disk suggestion, as I have a UPS connected anyway. If the system is shutdown via power management in response to low battery, will BOINC create a checkpoint?

I'm uncertain if I should enable hibernate on a desktop in OSX as I haven't needed it in years. (I loathe laptops.) Does anyone here have any input on how reliable hibernate is in OSX or if it will wake itself after power is restored?

What I did was I bought a good SSD for my OS and then a small 30Gb SSD for boinc.

I run boinc on this SSD even with CEP2 and have notched up just over 10 years of run time using a 6 core processor with no issues. Atleast then, I have the peace of mind that if there is every a problem, boinc won't contribute to it.