During a recent SETI outage recovery, I noticed a problem related to 'Redundant result - Cancelled by server' (server-side abort of work which has reached quorum and been validated before the host has even started computation - a new feature which I think only SETI has implemented so far).

When the servers came back up, my fast box was allocated a lot of WUs quite quickly, but struggled to download them through the congestion. However, other people obviously did download them, because some became redundant before I'd even managed to complete the download.

Once I'd reported the (aborted) result, my trusty BoincView was showing green across the board, but no new work was being fetched. I discovered that the machine was still re-trying five downloads, but getting 'file not found' (obviously, because the WU had been assimilated and the datapak deleted from the server). I just aborted the meaningless downloads, and work fetch started back up as normal.

BOINC needs to check for, and cancel, any associated transfers when doing a server-mandated abort. I'll log a Trac report. Edit - trac 366.