Additional information: My system's virtual memory (swap file) is turned off. I will not turn it on because I do not want sensitive stuff written there behind my back. However, I expect 2 GB of RAM to be plenty.

Off topic I know, but it does happen a lot. The pagefile (swap file, VM, etc...) and unused space on a hard drive can contain valuable data for extended periods of time. The less a hard drive is written to, the longer the chances are that the data will remain. So many people think that simply reformatting their hard drives wipes out the data...
If a person is worried about enabling swap files, flash drives can be used for this. I know it wears them out quicker, but with prices being dirt cheap for these things, a 1-4GB flash drive could be used as a swap file and taken out when the computer is not in use.
Again, sorry for the off-topic reply, just throwing in my 2 cents.

Back on topic, Tom, is there any reason why you don't want to run FAAH, HPF2, or DDDT? I have opted for the ACH,as well as the other three, but have yet to see a WU for ACH. You should check at least one other project so your machine will be utilized, not just idle and waiting for ACH work. There is the option to send work from other projects if no work is available for the one you checked, so that might work also. Those C2D machines can really push those work units out also, with very manageable processor temps, even at 100%.

Here a sample link to the default custom device profile: http://www.worldcommunitygrid.org/ms/device/v...iguration.do?name=Default

The option is called "Leave applications in memory while preempted?". Set that to Yes. In past it was helpful with multi-project clients, but the new scheduling even pauses one or the other WCG job to let another WCG job more urgent ahead. A previous project, HDC would keep the whole model of 1gb in RAM that way. That said, with VM switched off, BOINC has no option but to keep all in RAM. As for writing to flash drives and the like, timings are often too slow to work reliably. Errored results or aborts have been reported.

So is there a projected or estimated time frame for this project to be running again?

Past experience showed it kept the job in RAM at checkpoints when switching tasks, so resumption is quicker (remember my laments from early HDC days). With it off, it unloads. Large models would need time to restart fetching it from disk regardless if 5.8 and up switch at checkpoints. That latter was a major improvement to reduce loss of progress with preempting and not having LiM on. With large RAM it thus remains my recommendation.