Looks like a combination of reissues, machines that don't finish much at all and the known decade sensitivity for the work units that crashed on your machine. Since you have a Mac it might be worth stocking up on ANZ and EU models until a clearer picture emerges about the new HADCM3N batch.

Thanks Iain, I've deleted Hadcm3n from my choices & will keep an eye on future developments.
Out of interest, how does one check if a task is a reissue?

Looks like a combination of reissues, machines that don't finish much at all and the known decade sensitivity for the work units that crashed on your machine. Since you have a Mac it might be worth stocking up on ANZ and EU models until a clearer picture emerges about the new HADCM3N batch.

Thanks Iain, I've deleted Hadcm3n from my choices & will keep an eye on future developments.
Out of interest, how does one check if a task is a reissue?

Over the years there have been different categories of reissue:

1. Normal: The first model in a work unit might be called something like hadam3p_pnw_w1y6_2008_1_009351815_0 with the second hadam3p_pnw_w1y6_2008_1_009351815_1 etc. These are the usual failures followed by an immediate reissue.

2. Resubmission: Rarely, there have been reissues of existing work unit models with a new application version. Again, the final suffix increments but the model completion dates are very widely separated. Perhaps this happens when work units marked "no resubmission" are unmarked.

3. Zombie: There seems to be a problem in the BOINC software somewhere that causes work units marked as "no resubmission" to appear again - after a fixed time period. These are a disaster since the ancillary files have usually (and rightly) disappeared and all the models crash.

4. It is I, Leclerc: In this method models with the same parameters and ancillary files are reissued with a new name and possibly a new application - but the thin disguise is only that: they are the same models being retested.