Now that mine have run long enough, yes, it seems that they are estimated about twice heavier than they are. Maybe to be sure that the client starts them early enough for completing in time?

Note that the actual estimated duration that you see in the tasks list also depends on the current Duration Correction Factor of the client, which may be below or above reality depending on the other projects that you had run before.

CR@P!
missed it
2 machines twiddling their thumbs for hours and got nothing for it...

Yes, the servers know about the speed of a device and will than when needed give extra time for these monsters [which is standard 8 days for the A type]. This will have been a slower device that usually returns tasks within 2 days, but will not be able to do that with this work unit based on the projected hours.

The Linux quad picked one up last night, a first DDDT2 A type ever, with 53 hours (DCF 1.10) so put it ahead of the queue and watch how many will really be needed,

<ot> UPSs are really worth the money also because they [ours at least] also serve as spike protectors. Supposedly 70,000 joule these cheapos. You don't need the 1 hour version, just one that bridges short outages, particularly if you're in a reliable grid area. The APC we've had now for near 10 years recently had a new battery. Courier, 24 hours delivery from Milan, 32 Euro. It can tell the PC to orderly go into hibernation when it takes too long. </ot>

Good point. What came out of the feeder will likely still hold the last FPOP averages that the work generator knows from past production. Mine had 2:46 CPU hours at last checkpoint on exactly 20 percent (see 2 checkpoints on the quick) i.e. not 53 hours projected, but about 13-15 hours.