The three day limit on non-retry ARP1 tasks is quite normal for work units with low generation numbers (the three-digit number part of the name).

At some point in the past there were a lot of tasks that failed because of data conditions, and this is part of the attempt to get the laggards[1] caught up again once the reason for the problem had been identified.

As part of the fix, work unit generations are now categorized as "Normal", "Accelerated" or "Extreme" depending on how far behind the highest current generation they are. Extreme and Accelerated units get sent out with the three-day deadline in an effort to close the gap, and 3 tasks are sent out at once for Extreme units to try to avoid a retry being needed if one of the initial tasks misses deadline.

There are 35609 distinct cells being modelled, and around 95% of them are currently categorized as Normal, with the highest generation 138 at present (the final generation will be around 182 as each one is two days and a year is being modelled.) Currently there are about 100 in the Extreme category (generations up to 123 at present) and a couple of thousand in the Accelerated category (generations 124 to 128 at present). So both of your jobs happen to be "Extreme" :-)

Hope that makes sense, and helps somewhat...

Cheers - Al.

[1] Not all laggards are down to failed work units; a lot of ARP1 tasks end up "No Reply" or get aborted by the client because they didn't start by the deadline, and if that happens to several generations of a particular cell it will fall behind quite quickly...

Actually the usual is 6 days for Normal units with 3 days for their re-sends after the halfway point of their target.

Extremes and Accelerated have a 3 day target with 1.5 days for their re-sends after their halfway point.

Halving the deadline for re-sends is to help the re-sends to catch up with the initial distribution.

Al has explained about the differences between the categories.

0034097 does not conform to this as it doesn't seem to be a re-send.

Mike