Adri,

It is indeed a puzzle, Is it a client software thing or some sort of time error?

If that system is running a custom BOINC client, who knows what might happen :-) so I'm going to ignore that possibility!

I wonder if that client had a look at what to start next, decided it couldn't find anything that had a chance of getting finished in a timely fashion [or hit a resource limit that meant it could no longer start any of them anyway] and did a "Not Started by Deadline" on all of them?

Alternatively, what happens if one does a project reset with a load of work that can't sensibly be re-issued? Is the client clever enough to say "no point re-fetching any of these as they'd probably not run - kill them off as part of the reset..." (A normal project reset should just quietly re-fetch the tasks [I tested that on a system with almost no work :-)] and detaching should earn a Detached flag rather than an Error!)

Someone like Richard Haselgrove might know an answer to the above - I don't :-)

Then again, maybe it was a "perception of time" thing. Perhaps the client clock was out by several hours (returning from a system outage or similar), or perhaps this is one of those cases where the WCG server either told the client a different due time or [less likely] has recorded a bad sent time.

The above is [of course] speculation, but I hope it's informed speculation :-)

Cheers - Al.

P.S. It would be a lot easier to work out what was going on with this lot if the client hardware was a known quantity -- for instance, I wonder how long it would take it to successfully run a single SCC1 task?

I got MCM and SCC WUs on my Windows machine late last night but had to wait till early this morning to get them on my Linux rig.

Is there a known difference between the two systems?