Well, even with BOINC data-directory (or only slots-directories) on 2-disk RAID & OS on separate disk I've little problems generating "no heartbeat" if starts with any disk-heavy operations. Even if left computer alone the RAID-setup on an i7-920 with 6 GB ram running 8 CEP2 gave "no heartbeat" after a couple days.

Have also tried pre-starting CEP2, similar to Rickjb recommended, and while don't remember the exact number but AFAIK around 20 unzipped CEP2, hitting the disk-tab was enough to trigger "no heartbeat". This even happened after all CEP2 was removed from memory and was running some other work, clearly showing it's the number of files what's the problem and not how many GB they're using.

BTW, even without RAID, I've had no problems (*) running CPDN, even these also uncompresses around 350 MB of data on 1st. start of model. The big difference is CPDN uses only roughly 100 files, not 6466 like CEP2 does.

(*) I'm not counting problems due to buggy wu's and so on.

lolll...yes, I wandered "off-topic" if your intent was strictly to push the CEP2 project...well, really to push the corporation that wrote the analysis core (Q-Chem, which performs the quantum mechanics calculations) of this phase - 2 - of the CEP project to change their code...

Well, from my personal experience with both writing code and with dealing with other corporations' software (and their lead product, at that)...I would venture that is going to be tough to get done. Unless you pony up...oh, say half a million dollars - at a guess. (But I don't reckon either Q-Chem or Harvard would be averse to the idea...assuming the rewrite could be completed, tested, and rolled out while there was still sufficient work units left to make it worthwhile.)

My apologies for "wandering a bit OT"...I tend to focus upon the pragmatic...the deliverables.

(Edit: I'd add that I was addressing RAID or at least a separate data drive allows the multithreading capabilities of modern operating systems [Windows et al] to avoid having to queue I/O in a single thread. In the case of I/O contention, it doesn't matter if you're running just CEP2 or CEP2 and other sciences - if they demand I/O at the same time as another science or at the same time as the operating system, you get I/O collisions. Said collisions will, of course, be exacerbated if you have a single science starting out from 0% across multiple cores...which I don't have any issues dealing with using separate RAID sets for the system and data drives.)

Dear ibsteve2u et al,
thanks so much for your enthusiasm - we greatly appreciate it! We are very grateful for all the work you and all the WCG crunchers put into our project. We are definitely trying to attract more people to the project. We would, however, resist from ranking the value of the different research projects. They are all great and they all work towards the common good. We think that people should support whatever science they personally relate to - if that ends up being CEP that's awesome, if it's one of the other projects, that's great, too. It's not a competition for us.
Dear Rickjb,
yes, improving the checkpointing is still on our TODO list, but that list is so long and our manpower is so limited that we just have to prioritize, and there are other things that are just more pressing than the checkpointing.
Best wishes from
Your Harvard CEP team

My apologies; I will temper my efforts with your more...neutral...perspective.

I've been running the app config on 3 computers for a week or so and it works great for this not having to micromanage all the CEP WU's. Thank you Sekerob as always for all your awesome charts, especially the pie chart with all the projects.