Work aborted by user

The project only sends work when the BOINC client requests it. So the question you need to ask is, "Why did the BOINC client request work beyond the deadline?" Obviously, at the time BOINC requested that work, BOINC didn't think the deadline would be exceeded. Something must have changed to invalidate BOINC's assumptions.

The answer may be complex and it depends on a large number of factors, such as your mix of projects, their various resource shares, the size of the work cache you try to maintain, the various apps supplied by these projects, what other non-BOINC compute intensive jobs may have run, have any projects had work supply issues recently, how close to reality are the time estimates for tasks when you first receive them, are there work fetch bugs in the version of BOINC you are currently using, etc, etc?

The vast majority of these things are not under the control of any particular project so you do have to be careful with your preference choices. If you have a couple of projects and perhaps a couple of apps per project, BOINC can get itself in a bind very easily if you try to maintain a large cache size, for example.

Unfortunately, only you can see the details needed to work out what may have happened. If you'd like to give more information, it would be easier to pinpoint the likely problem. If I had to guess, from a quick inspection of your tasks on the website, I'd say your work cache settings are too large. Prior to the problem, there are tasks showing with around 12-14 days between time of sending and time of return. It wouldn't take much of a disturbance to push a bunch of tasks beyond the deadline.