As my colleagues have said, this is a rare problem now. HPF2 suffered quite badly from it at one point.

More statistics would be useful. Are we seeing an increase in this problem, or is it just a coincidental report?

I would be interested to know whether the affected computers change power modes frequently. Do they hibernate, or go into sleep mode?

As you can imagine, this is categorised as a "hard" bug. We can't reproduce it on demand, and often it doesn't show up at all. This makes debugging very difficult.

My previous answer to this statement was more about the effect of Boinc offloading a WU completely. Regarding the amount of RAM and disk I/O now.

If you have enough RAM why not leave a temporarily inactive task in memory anyway? If RAM is really enough nothing particular will happen (my own case). If at some point some more RAM is necessary the system itself will take what is necessary (and not more) and when the WU will resume it will only have to reload those pages which had been dropped to the swap file. At worst that will be the same amount of disk I/O if it had to take all the memory space of the suspended task.

In fact excessive disk I/O because of insufficient RAM occurs only when active tasks need more RAM than available. Inactive tasks do not matter in this case. They have an effect only on the size of the necessary swap file.

Cheers. Jean.

That's not entirely accurate. I'll ask Jord about it and see if we can come up with a more accurate description.

I have done a test this evening with Boinc 5.10.45 under XP.
I have changed my local prefs to LIM OFF (unchecked the box), then I started TaskManager (TM) sorted on the process names.
When I snoozed Boinc from the SysTray icon the four DDDT WUs disappeared from TM. When I "unsnoozed" they resumed at their latest checkpoints.
Same result when I suspended Boinc from the Activity submenu of the Advanced view.
Same result again when I suspended WCG from the project tab.

I finally checked the box "Leave in memory..." again (although I never crunched for anything else than WCG), and now it is not just "because that does not hurt", it is because it is what it should be to minimize losses.

I had no repair job at this time. Next time I have one I shall check if the interrupted WU is offloaded when the high priority task starts. Probably yes, but it is better to make sure with a test

Cheers. Jean..

You are a little severe with Jord.
He is clearly wrong once about suspending the project wiith LIM ON, but he is only too talkative about what the system does with LIM ON. If all RAM is really needed by active processes the suspended task will really been kept "in memory", but only in the pagefile finally. But giving those details is probably only adding confusion for some people.

Here I think you are very indulgent, unless I have not found the right documentation.
The official wiki talks about applications "preempted" without a single explanation about what preemption is.
The unofficial wiki is not much better about "Leave in memory" but it has an article about preemption. And it is because I have reread it yesterday that I decided to run these tests. Because after reading this article my understanding was that preemption applied only to project switching, and not to any of the other circumstances where a task is suspended or interrupted.
Either this is true in Boinc terminology and the description of LIM is wrong since it does not apply only when switching projects.
Or preemption does not apply only to project switching and the descriptions of LIM here or there are not only cryptic but confusing too.

After testing/checking a few more things in the coming days I think I will go on with my intention to write my first FAQ about LIM, with normal words for normal people, hopefully.

Cheers. Jean.